CN111981635A - Central air conditioner fault prediction and diagnosis method adopting double intelligent algorithms - Google Patents

Central air conditioner fault prediction and diagnosis method adopting double intelligent algorithms Download PDF

Info

Publication number
CN111981635A
CN111981635A CN202010702092.3A CN202010702092A CN111981635A CN 111981635 A CN111981635 A CN 111981635A CN 202010702092 A CN202010702092 A CN 202010702092A CN 111981635 A CN111981635 A CN 111981635A
Authority
CN
China
Prior art keywords
neuron
central air
representing
diagnosis
air conditioner
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010702092.3A
Other languages
Chinese (zh)
Other versions
CN111981635B (en
Inventor
何新
张博譞
张秋实
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenyang Anxin Automation Control Co ltd
Original Assignee
Shenyang Anxin Automation Control Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenyang Anxin Automation Control Co ltd filed Critical Shenyang Anxin Automation Control Co ltd
Priority to CN202010702092.3A priority Critical patent/CN111981635B/en
Publication of CN111981635A publication Critical patent/CN111981635A/en
Application granted granted Critical
Publication of CN111981635B publication Critical patent/CN111981635B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F11/00—Control or safety arrangements
    • F24F11/30—Control or safety arrangements for purposes related to the operation of the system, e.g. for safety or monitoring
    • F24F11/32—Responding to malfunctions or emergencies
    • F24F11/38—Failure diagnosis
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F11/00—Control or safety arrangements
    • F24F11/62—Control or safety arrangements characterised by the type of control or by internal processing, e.g. using fuzzy logic, adaptive control or estimation of values
    • F24F11/63—Electronic processing
    • F24F11/64—Electronic processing using pre-stored data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Signal Processing (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Computing Systems (AREA)
  • Mechanical Engineering (AREA)
  • Combustion & Propulsion (AREA)
  • Chemical & Material Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Fuzzy Systems (AREA)
  • Testing And Monitoring For Control Systems (AREA)
  • Air Conditioning Control Device (AREA)

Abstract

A central air conditioner fault prediction and diagnosis method based on a double intelligent algorithm comprises the following steps: collecting a real-time data set of the operation of the on-site central air conditioner; inputting the real-time data set in the step one into a pre-established central air conditioner fault prediction and diagnosis classification neural network model based on a simulated annealing algorithm and an extreme learning machine algorithm to judge whether the real-time data set in the step one meets the condition that the central air conditioner fault prediction and diagnosis misclassification rate are minimum; and if the real-time data set in the step two meets the condition that the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, outputting the fault of the central air conditioner. And (4) predicting and diagnosing classification results. The method has field feasibility and is convenient for engineering application. Meanwhile, the extreme learning machine has good generalization capability and higher operation efficiency. Therefore, the problem of minimum misclassification rate of the central air conditioner is solved by adopting a simulated annealing algorithm, so that a global optimal solution is easier to obtain, and the accuracy and precision of fault detection and diagnosis model classification of the on-line deployed central air conditioner are improved.

Description

Central air conditioner fault prediction and diagnosis method adopting double intelligent algorithms
The technical field is as follows:
the invention relates to the technical field of central air conditioner fault prediction and diagnosis, in particular to a double-intelligent-algorithm central air conditioner fault prediction and diagnosis method based on a simulated annealing algorithm and an extreme learning machine algorithm.
Background art:
the central air-conditioning system is not only an indispensable energy consumption operation system in modern buildings, but also an important component of a building automation control system. Along with the improvement of the requirements of people on the living quality, the design of a central air conditioning system is more and more complex, and the central air conditioning system not only can provide refrigeration or heating, but also can provide domestic hot water, bath water, constant-temperature hot water for a swimming pool and the like. If the central air conditioner fails, inconvenience is brought to the operation of the system and the life of people, and meanwhile, the energy consumption of the system is increased by 15% -30% due to the failure of the central air conditioner. Therefore, the fault prediction and diagnosis technology based on the machine learning intelligent algorithm is applied to the central air-conditioning system, accurate and rapid fault parts are judged, the equipment maintenance time is shortened, the energy consumption of air-conditioning equipment is reduced, and the method has extremely important practical significance.
In the traditional method for predicting and diagnosing the faults of the central air conditioner, frequency domain signals of the central air conditioner system, which are collected by laboratory equipment, are used as characteristic vectors of a fault prediction and diagnosis model. However, the laboratory frequency domain signal acquisition and analysis equipment has high sensitivity, and the operation site of the central air-conditioning system is difficult to acquire and analyze the site frequency domain signals due to environmental constraints, so that the central air-conditioning fault prediction and diagnosis model established according to the signals acquired and analyzed by the laboratory equipment is difficult to deploy and implement on the industrial site. The existing linear regression fault prediction and diagnosis model cannot identify the nonlinear characteristics of the faults of the central air conditioner, and the prediction and diagnosis accuracy is poor. The central air conditioner fault prediction and diagnosis model of the back propagation neural network has the advantages of complex network structure, long training time and low feasibility of on-line deployment and on-line detection. The problem of accuracy of the central air conditioner fault prediction and diagnosis model can be solved by converting the problem into an optimization problem with the minimum fault misclassification rate through a mathematical modeling method. Each new solution of the genetic algorithm needs to be subjected to processes of coding, crossing, mutation and the like, the calculated amount is large, and the precision of the optimized solution is low. The hill climbing algorithm only receives the optimal solution in each iteration process, and is easy to fall into the local optimal solution.
Disclosure of Invention
The purpose of the invention is as follows: the invention provides a double-intelligent-algorithm central air conditioner fault prediction and diagnosis method based on a simulated annealing algorithm and an extreme learning machine algorithm, and aims to predict and diagnose faults of a central air conditioner by using time domain signal data acquired by a central air conditioner field sensor, improve the fault detection accuracy of the central air conditioner, determine the accurate position of the fault, finally improve the working efficiency of a system, reduce the energy consumption of the system and ensure the safe operation of the system.
The technical scheme is as follows: the invention is realized by the following technical scheme:
a double-intelligent-algorithm central air conditioner fault prediction and diagnosis method is characterized by comprising the following steps: the method comprises the following steps:
firstly, collecting a real-time data set S for the operation of a central air conditioner on siteon-line；
Secondly, the real-time data set S in the step of one is collectedon-lineInputting the data into a pre-established central air conditioner fault prediction and diagnosis classification neural network model based on a simulated annealing algorithm and an extreme learning machine algorithm to judge the real-time data set S in the step oneon-lineWhether the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum or not is met;
real-time data set S in the third and second stepson-lineAnd if the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, outputting a fault prediction and diagnosis classification result of the central air conditioner.
In the step two: if the real-time data set S in the step of "oneon-lineIf the fault prediction and diagnosis misclassification rate of the central air conditioner is not met to be minimum, the field central air conditioner operation data set S is usedon-lineAnd establishment ofInitial data set S collected during central air conditioner fault prediction and diagnosis classification modeloff-lineAnd merging, and training a central air conditioner fault prediction and diagnosis classification neural network model.
The method for constructing the central air-conditioning fault prediction and diagnosis classification neural network model comprises the following steps:
step 1: the historical operation data collected by the central air-conditioning heating/cooling heating and cooling cycle field sensor is used as an initial data set Soff-lineAnd the selected initial data set Soff-lineThe data comprises data with normal operation and data with failure operation;
step 2: judging whether the fault prediction and diagnosis classification neural network model meets the condition that the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum;
and step 3: if the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, entering a step 7;
and 4, step 4: if the fault prediction and diagnosis misclassification rate of the central air conditioner is not met, entering a step 6;
and 5: optimizing the central air-conditioning fault prediction and diagnosis classification model coefficient of the extreme learning machine by adopting a simulated annealing algorithm to obtain a new central air-conditioning fault prediction and diagnosis classification neural network model, and then judging whether the new central air-conditioning fault prediction and diagnosis classification neural network model meets the condition that the central air-conditioning fault prediction and diagnosis misclassification rate is minimum; if the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, entering
Step 6; otherwise, continuing optimization until the minimum misclassification rate is met;
and 7: and deploying a central air conditioner fault prediction and diagnosis classification model of a double intelligent algorithm on line.
In step 2, the method for training the central air-conditioning fault prediction and diagnosis classification neural network model of the extreme learning machine specifically comprises the following steps:
step 1: the historical operation data collected by the central air-conditioning heating/cooling heating and cooling cycle field sensor is used as an initial data set Soff-lineAnd the selected sample comprises a fortuneNormal data and operation fault data;
step 2: constructing a single-layer feedforward neural network model consisting of 12 input neurons, 4 output neurons and L hidden layer neurons; wherein the input neurons are respectively at inhalation pressure x1Evaporator approach temperature x2The approach temperature x of the condenser3Return water temperature x of evaporator4Outlet water temperature x of evaporator5Condenser return water temperature x6Outlet water temperature x of condenser7Water supply pressure x of water separator8Water collector return pressure x9Current x10Voltage x11Heating/cooling and refrigerating capacity x of central air conditioner12(ii) a The output neurons are respectively of the normal operation type y of the central air conditioner1Evaporator failure type y2Type of condenser failure y3And compressor failure type y4；
And step 3: normalizing the 12 input neurons;
and 4, step 4: initial data set Soff-lineDivided into training sample sets StrainAnd a test sample set Stest(ii) a Wherein the training sample set StrainThe data comprises data of a normal operation type and data of an operation fault type, and the total number of the N discrete training samples
Figure BDA0002593168850000031
Wherein xi＝[xi1,xi2,…,xi12]T∈R12(ii) a Component xijA j-th input neuron representing an i-th sample, and i ═ 1,2, …, N; j ═ 1,2, …, 12; x is the number ofi11 st input neuron inspiratory pressure, x, representing the ith samplei2The 2 nd input neuron evaporator approach temperature, x, representing the ith samplei3The 3 rd input neuron condenser approach temperature, x, representing the ith samplei44 th input neuron evaporator Return Water temperature, x, representing the ith samplei5The evaporator leaving water temperature, x, of the 5 th input neuron representing the ith samplei6Condenser backwater for the 6 th input neuron representing the ith sampleTemperature, xi7Condenser outlet water temperature, x, of the 7 th input neuron representing the ith samplei8Water supply pressure, x, for the 8 th input neuron trap representing the ith samplei9The 9 th input neuron collector return water pressure, x, representing the ith samplei1010 th input neuron current, x, representing the ith samplei11The 11 th input neuron voltage, x, representing the ith samplei12The 12 th input neuron central air-conditioning heating/cooling heating and cooling capacity representing the ith sample; x is the number ofiAs an input to a single-layer feedforward neural network model; etai＝[ηi1,ηi2,ηi3,ηi4]T∈R4Representing normal operation result and fault operation result data contained in the sample; component ηi1The ith sample is the normal operation type of the central air conditioner, etai2Indicating the i-th sample as evaporator fault type, etai3Indicating the ith sample as the type of condenser fault, ηi4Indicating the ith sample as a compressor fault type; symbol [ 2 ]]TRepresents a transpose of a matrix;
and 5: selecting L as the number of hidden layer neurons;
step 6: adopting a random method to generate the weight value w of the kth neuron of the hidden layer and the 12 neurons of the input layerk＝[wk1,wk2,…,wk12]T(ii) a Generating a bias value b of the kth neuron of the hidden layer by adopting a random methodk，k＝1,2,…,L；
And 7: selecting an excitation function g (w) of a central air-conditioning fault prediction and classification neural network model of the extreme learning machinek·xi+bk) And i is 1,2, …, N, k is 1,2, …, L;
and 8: using the weight value w calculated in step 61,…,wLAnd an offset value b1,...,bLAnd the excitation function in the step 7, and calculating the hidden layer output matrix H ═ H (w)1,...,wL,b1,...,bL,x1,...,xN)；
And step 9: using results from step 8A hidden layer output matrix H, calculating the weight beta from L neurons of the hidden layer to neurons of the output layer [ beta ]1,β2,…,βL]T(ii) a Wherein beta isk＝[βk1,βk2,βk3,βk4]TThe weight k representing the k-th neuron of the hidden layer to 4 neurons of the output layer is 1,2, …, L;
step 10: judging whether the beta in the step 9 meets the requirement of minimum output error of the single-layer feedforward neural network, if so, entering a step 11; otherwise, returning to the step 5;
step 11: outputting a central air conditioner fault prediction and classification model based on an extreme learning machine;
step 12: judging whether the central air-conditioning fault prediction and classification model based on the extreme learning machine in the step 11 meets the condition that the central air-conditioning fault prediction and diagnosis misclassification rate is minimum; if yes, entering step 14, otherwise, directly entering step 13;
step 13: if the model does not meet the condition that the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, returning to the step 9, and optimizing the weight beta from L neurons of the hidden layer to neurons of the output layer by adopting a simulated annealing algorithm [ beta ]1,β2,…,βL]T；
Step 14: verifying the trained central air conditioner fault prediction and classification model by using the test sample;
step 15: judging whether the test sample meets the condition that the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum; if yes, go to step 16; otherwise, returning to the step 5.
Step 13, optimizing the weights beta from L neurons in the hidden layer to neurons in the output layer by adopting a simulated annealing algorithm1,β2,…,βL]TThe method comprises the following specific steps:
step 1: initializing relevant parameters of a simulated annealing algorithm, and selecting any initial solution beta in a feasible solution space based on a central air conditioner fault prediction and diagnosis model of an extreme learning machine;
step 2: calculating the misclassification rate W of the central air-conditioning fault prediction and diagnosis model generated by betaerror(β)；
And step 3: randomly generating beta*Calculating by beta*Misclassification rate W of generated central air conditioner fault prediction and diagnosis modelerror(β*)；
And 4, step 4: judgment of Werror(β*)-Werror(β) is less than or equal to 0;
and 5: if Werror(β*)-WerrorIf beta is less than or equal to 0, then beta is used*Generating a central air conditioner fault prediction and diagnosis model, and entering the step 7;
step 6: if Werror(β*)-Werror(beta) > 0, then a bit at [0,1 ] is randomly generated]Random number in interval
Figure BDA0002593168850000041
Judgment of
Figure BDA0002593168850000042
Whether or not greater than
Figure BDA0002593168850000043
If it is
Figure BDA0002593168850000044
Then by beta*Generating a central air conditioner fault prediction and diagnosis classification model, and entering the step 7; if it is
Figure BDA0002593168850000045
The parameters of the central air conditioner fault prediction and diagnosis classification model are unchanged; the symbol exp () represents the exponent calculation;
and 7: judging whether the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum; if yes, entering step 8; otherwise, returning to the step 1, and reducing the temperature control parameter t to continue the operation.
In step 2, the misclassification rate W of the central air conditioner fault prediction and diagnosis model generated by beta is calculatederrorThe method (β) is specifically as follows:
the quality evaluation of the central air-conditioning fault prediction and diagnosis classification model is evaluated by adopting fault misclassification rate conversion, and a misclassification rate calculation method is described by taking a training sample set as an example;
training sample set StrainThe data comprises normal operation data and operation fault data, and the total number of the N discrete training samples
Figure BDA0002593168850000051
W1The method comprises the steps of misclassifying a sample, which is output as a normal operation type by a central air conditioner fault prediction and diagnosis model, as a fault type, or misclassifying a sample, which is output as a fault type by a model, as a set of normal operation types;
W2the method comprises the steps of classifying a sample which is output as a fault state by a model into a fault type, but not belonging to a data set of a correct fault type sample; if the original output type is evaporator fault but misclassification is condenser fault or compressor fault, the original output type is condenser fault but misclassification is evaporator fault or compressor fault, and the original output type is compressor fault but misclassification is evaporator fault or condenser fault;
W3is a set of classification rates for which the output type is correct, and satisfies
Figure BDA0002593168850000054
Wherein the symbols
Figure BDA0002593168850000055
Is a direct sum operation symbol;
the minimum misclassification transformation of the central air-conditioning fault prediction and diagnosis is as follows:
Figure BDA0002593168850000052
Figure BDA0002593168850000053
in which ξij(beta) represents a central air conditioner fault prediction and diagnosis classification model classification function by a dual intelligent algorithm; y isiRepresenting output layer neurons trained by neural networks, yi＝[yi1,yi2,yi3,yi4]TI is 1,2, …, N, and satisfies the output error min yi-ηi||。
In step 3, the 12 input neuron parameters are normalized to 0-1 numerical values, and the specific normalization method is as follows:
(1) inspiratory pressure parameter x1Is normalized to obtain
Figure BDA0002593168850000061
Figure BDA0002593168850000062
Wherein
Figure BDA0002593168850000063
The air suction pressure under the working condition is designed for the central air-conditioning system,
Figure BDA0002593168850000064
is a normalized inspiratory pressure parameter;
(2) evaporator approach temperature parameter x2Is normalized to obtain
Figure BDA0002593168850000065
Figure BDA0002593168850000066
Wherein
Figure BDA0002593168850000067
The evaporator approaches the temperature under the design working condition of the central air-conditioning system,
Figure BDA0002593168850000068
the evaporator approach temperature parameter after normalization;
(3) condenser approach temperature x3Is normalized to obtain
Figure BDA0002593168850000069
Figure BDA00025931688500000610
Wherein
Figure BDA00025931688500000611
The approach temperature of the condenser under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500000612
the normalized condenser approach temperature parameter is obtained;
(4) return water temperature x of evaporator4Is normalized to obtain
Figure BDA00025931688500000613
Figure BDA00025931688500000614
Wherein
Figure BDA00025931688500000615
The return water temperature of the evaporator under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500000616
the normalized return water temperature parameter of the evaporator is obtained;
(5) water outlet temperature x of evaporator5Is normalized to obtain
Figure BDA00025931688500000617
Figure BDA00025931688500000618
Wherein
Figure BDA00025931688500000619
The outlet water temperature of the evaporator under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500000620
the normalized outlet water temperature parameter of the evaporator is obtained;
(6) condenser return water temperature x6Is normalized to obtain
Figure BDA00025931688500000621
Figure BDA00025931688500000622
Wherein
Figure BDA00025931688500000623
The return water temperature of the condenser under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500000624
the normalized condenser return water temperature parameter is obtained;
(7) condenser outlet water temperature x7Is normalized to obtain
Figure BDA00025931688500000625
Figure BDA00025931688500000626
Wherein
Figure BDA00025931688500000627
Designing the outlet water temperature of the condenser under the working condition for the central air-conditioning system;
Figure BDA00025931688500000628
the water outlet temperature parameter is the normalized condenser water outlet temperature parameter;
(8) water supply pressure x of water separator8Is normalized to obtain
Figure BDA00025931688500000629
Figure BDA00025931688500000630
Wherein
Figure BDA00025931688500000631
The water supply pressure of the water distributor under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500000632
the water supply pressure parameter of the water separator after normalization;
(9) water collector return pressure x9Is normalized to obtain
Figure BDA00025931688500000633
Figure BDA00025931688500000634
Wherein
Figure BDA00025931688500000635
The backwater pressure of the water collector under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500000636
the water pressure parameter is the normalized water collector return water pressure parameter;
(10) current x10Is normalized to obtain
Figure BDA00025931688500000637
Figure BDA00025931688500000638
Wherein
Figure BDA00025931688500000639
Rated current for safe operation of the central air-conditioning system;
Figure BDA00025931688500000640
is a normalized current parameter;
(11) voltage x11Is normalized to obtain
Figure BDA0002593168850000071
Figure BDA0002593168850000072
Wherein
Figure BDA0002593168850000073
For central air-conditioning systemRated voltage for full operation;
Figure BDA0002593168850000074
the normalized voltage parameter is obtained;
(12) heating and cooling capacity x of central air conditioner12Is normalized to obtain
Figure BDA0002593168850000075
Figure BDA0002593168850000076
Wherein
Figure BDA0002593168850000077
The calculation method is as follows for the nominal heating capacity of the central air conditioner:
Figure BDA0002593168850000078
wherein K1Correcting coefficient of heating capacity or cooling capacity of the central air conditioning unit; k2A defrosting correction coefficient of the central air conditioning unit; k, the comprehensive heat transfer coefficient of the building; t isnHeating and cooling indoor design temperature, TpCalculating the outdoor temperature in the heating season or the refrigerating season;
Figure BDA0002593168850000079
heating and refrigerating capacity parameters of the normalized central air conditioner;
where the symbol | | | represents the modulus of the vector.
Step 8 calculates the hidden layer output matrix H ═ H (w)1,...,wL,b1,...,bL,x1,...,xN)：
Figure BDA00025931688500000710
Wherein x1＝[x11,x12,…,x112]TRepresents the 1 st sample parameter in the training dataset: suction pressure, evaporator approach temperature, condenser approach temperature, evaporator return water temperature, evaporator outlet water temperature,The water supply pressure of the water separator, the water return pressure of the water collector, the current, the voltage and the heating and refrigerating capacity of the central air conditioner are measured; the 12 parameters are used as input layer parameters of a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
wherein xN＝[xN1,xN2,…,xN12]TRepresents the nth sample parameter in the training dataset: the air-conditioning system comprises an air suction pressure, an evaporator approach temperature, a condenser approach temperature, an evaporator return water temperature, an evaporator outlet water temperature, a condenser return water temperature, a condenser outlet water temperature, a water distributor supply pressure, a water collector return water pressure, current, voltage and central air-conditioning heating and refrigerating capacity; the 12 parameters are used as input layer parameters of a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
wherein w1＝[w11,w12,…,w112]Representing the weight values of the 1 st neuron of the hidden layer and the 12 neurons of the input layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm, and the component w11Representing the weight value of the 1 st neuron of the hidden layer and the inspiration pressure neuron of the input layer; w is a12Representing the weight value of the 1 st neuron of the hidden layer and the approximate temperature neuron of the evaporator of the input layer; w is a13Representing the weight value of the 1 st neuron of the hidden layer and the input layer condenser approaching temperature neuron; w is a14Representing the weight values of the 1 st neuron of the hidden layer and the return water temperature neuron of the evaporator of the input layer; w is a15Representing the weight values of the 1 st neuron of the hidden layer and the water outlet temperature neuron of the input layer evaporator; w is a16Representing the weight values of the 1 st neuron of the hidden layer and the return water temperature neuron of the condenser of the input layer; w is a17Representing the weight values of the 1 st neuron of the hidden layer and the water outlet temperature neuron of the input layer condenser; w is a18Representing the weight values of the 1 st neuron of the hidden layer and the water supply pressure neuron of the input layer water separator; w is a19Representing the weight value of the 1 st neuron of the hidden layer and the backwater pressure neuron of the water collector of the input layer; w is a110Representing the 1 st neuron and input of the hidden layerA weight value of the in-layer current neuron; w is a111Representing the weight values of the 1 st neuron of the hidden layer and the input layer voltage neuron; w is a112Representing the weight values of the 1 st neuron of the hidden layer and the heating and cooling capacity neuron of the central air conditioner of the input layer;
wherein wL＝[wL1,wL2,…,wL12]Representing the weight values of the Lth neuron of the hidden layer and the 12 neurons of the input layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm, and the component wL1Representing the weight value of the Lth neuron of the hidden layer and the inspiration pressure neuron of the input layer; w is aL2The weight value of the Lth neuron of the hidden layer and the approximate temperature neuron of the evaporator of the input layer is represented; w is aL3Representing the weight value of the Lth neuron of the hidden layer and the input layer condenser approaching temperature neuron; w is aL4Representing the weight values of the Lth neuron of the hidden layer and the return water temperature neuron of the evaporator of the input layer; w is aL5Representing the weight values of the L-th neuron of the hidden layer and the water outlet temperature neuron of the input layer evaporator; w is aL6Representing the weight values of the Lth neuron of the hidden layer and the return water temperature neuron of the condenser of the input layer; w is aL7Representing the weight values of the Lth neuron of the hidden layer and the water outlet temperature neuron of the input layer condenser; w is aL8Representing the weight value of the L-th neuron of the hidden layer and the water supply pressure neuron of the input layer water separator; w is aL9Representing the weight value of the Lth neuron of the hidden layer and the backwater pressure neuron of the water collector of the input layer; w is aL10Representing the weight value of the 1 st neuron of the hidden layer and the current neuron of the input layer; w is aL11Representing the weight values of the Lth neuron of the hidden layer and the voltage neuron of the input layer; w is aL12Representing the weight values of the Lth neuron of the hidden layer and the heating and cooling capacity neuron of the central air conditioner of the input layer; "·" denotes an inner product operation; with wL·x1The description is given for the sake of example: w is aL·x1＝wL1×x11+wL2×x12+…wL12×x112∈R；
Excitation function g (w)k·xi+bk) Center to be based on extreme learning machine algorithmsMapping a linear space containing suction pressure, evaporator approach temperature, condenser approach temperature, evaporator return water temperature, evaporator outlet water temperature, condenser return water temperature, water distributor supply pressure, water collector return water pressure, current, voltage and central air conditioner heating and refrigerating capacity of 12 input layer neurons in an air conditioner fault prediction and diagnosis neural network model into a linear space containing L hidden layer characteristics in an extreme learning machine; the specific selection method of the excitation function is as follows:
trigonometric function: g (w)k·xi+bk)＝cos(wk·xi+bk)；
Gaussian function:
Figure BDA0002593168850000081
sigmoid function:
Figure BDA0002593168850000082
hyperbolic sine function:
Figure BDA0002593168850000091
radial basis function: g (wk. xi + b)k)＝exp(-bk·xi-wk||)。
And step 9: calculating the weight hidden layer output weight beta from the hidden layer L neurons to the output layer neurons according to beta H + Y1,β2,…,βL]；
Wherein beta is1＝[β11,β12,β13,β14]TMiddle component beta11The central air conditioner fault prediction and diagnosis neural network model based on the extreme learning machine algorithm is characterized by comprising 1 st neuron and 1 st output layer neuron of a hidden layer, wherein the 1 st neuron and the 1 st output layer neuron are respectively connected with a central air conditioner; component beta12Representing the weight values of the evaporator fault types of the 1 st neuron and the 2 nd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta13Representing the weight values of the condenser fault types of the 1 st neuron and the 3 rd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta14Representing the weight values of compressor fault types of 1 st neuron and 4 th output layer neuron of a hidden layer in a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
wherein beta is2＝[β21,β22,β23,β24]TMiddle component beta21The central air conditioner fault prediction and diagnosis neural network model based on the extreme learning machine algorithm is characterized by comprising the following steps of (1) representing the weight value of the normal operation type of the central air conditioner of the hidden layer 2 th neuron and the hidden layer 1 st output layer neuron; component beta22Representing the weight values of the evaporator fault types of the neurons of the No. 2 and the No. 2 output layers of the hidden layers in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta23Representing the weight values of the condenser fault types of the 2 nd neuron and the 3 rd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta24Representing the weight values of compressor fault types of a hidden layer 2 nd neuron and a hidden layer 4 th output layer neuron in a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
wherein beta isL＝[βL1,βL2,βL3,βL4]TMiddle component beta11The central air conditioner fault prediction and diagnosis neural network model based on the extreme learning machine algorithm is characterized by comprising the following steps of (1) representing the weight value of the normal operation type of the central air conditioner of the hidden layer Lth neuron and the 1 st output layer neuron; component beta12Representing the weight values of the evaporator fault types of the L-th neuron and the 2 nd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta13Representing condenser of Lth neuron and 3 rd output layer neuron of hidden layer in central air-conditioning fault prediction and diagnosis neural network model based on extreme learning machine algorithmA weight value of the barrier type; component beta14Representing the weight values of compressor fault types of the Lth neuron and the 4 th output layer neuron of a hidden layer in a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
H+output matrix H ═ H (w) for hidden layers1,...,wL,b1,...,bL,x1,…,xN) Moore-Penrose generalized inverse matrix of (1); when H is presentTH is not singular: h+＝(HTH)-1HTOr H+＝HT(HTH)-1(ii) a Wherein ()-1Represents the inverse of the matrix ()TRepresents a transpose of a matrix;
Y＝[y1,y2,…,yN]Tzero-error approximation of N discrete training samples for training
Figure BDA0002593168850000101
Of the neural network, wherein yi＝[yi1,yi2,yi3,yi4]T，i＝1,2,…,N；
At y1Middle, component y11Representing that the 1 st neuron output after the training of the neural network model of the extreme learning machine is carried out on the 1 st sample by zero error approximation is the normal operation type of the central air conditioner; component y12Representing that the 2 nd neuron output after the 1 st training sample is subjected to extreme learning machine neural network model training is an evaporator fault type by zero error approximation; component y13Representing that the 3 rd neuron output after the training of the neural network model of the extreme learning machine is carried out on the 1 st training sample by using zero error approximation is a condenser fault type; component y14Representing that the 4 th neuron output after the training of the neural network model of the extreme learning machine is carried out on the 1 st training sample by using zero error approximation is a compressor fault type;
at y2Middle, component y21Representing that the 1 st neuron output after the 2 nd sample is subjected to the neural network model training of the extreme learning machine is the central air conditioner alignment by zero error approximationA normal operation type; component y22Representing that the 2 nd neuron output after the training of the neural network model of the extreme learning machine is carried out on the 2 nd training sample by zero error approximation is an evaporator fault type; component y23Representing that the 3 rd neuron output after the 2 nd training sample is subjected to extreme learning machine neural network model training is a condenser fault type by zero error approximation; component y24Representing that the 4 th neuron output after the 2 nd training sample is subjected to extreme learning machine neural network model training is a compressor fault type by zero error approximation;
at yNMiddle, component yN1Representing that the 1 st neuron output after the training of the neural network model of the extreme learning machine is carried out on the Nth training sample by using zero error approximation is the normal operation type of the central air conditioner; component yN2Representing that the Nth neuron output after the Nth training sample is subjected to extreme learning machine neural network model training by using zero error approximation is an evaporator fault type; component yN3Representing that the 3 rd neuron output after the training of the neural network model of the extreme learning machine is carried out on the Nth training sample by using zero error approximation is a condenser fault type; component yN4Representing that the 4 th neuron output after the training of the neural network model of the extreme learning machine is carried out on the Nth training sample by using zero error approximation is a compressor fault type;
yithe calculation method is
Figure BDA0002593168850000102
And the minimum min y of the error output by the neural network model is satisfiedi-ηi||，i＝1,2,…,N，k＝1,2,…,L。
The double-intelligent-algorithm central air-conditioning fault prediction and diagnosis system is characterized in that: the system comprises a data acquisition module, a data judgment module and a fault prediction and diagnosis type data output module:
the data acquisition module acquires a real-time data set S for the operation of the central air conditioner on siteon-line；
A data judgment module for collecting the real-time data set S in the data acquisition moduleon-lineInputting the data into a pre-established central air conditioner fault prediction and diagnosis classification neural network model based on a simulated annealing algorithm and an extreme learning machine algorithm to judge whether the real-time data set in the step one meets the condition that the central air conditioner fault prediction and diagnosis misclassification rate is minimum;
the fault prediction and diagnosis type data output module: real-time data set S in data judgment moduleon-lineAnd outputting the central air conditioner fault prediction and diagnosis classification result if the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum.
The advantages and effects are as follows:
the invention designs a method for predicting and diagnosing faults of a central air conditioner by using time domain signal data acquired by a central air conditioner field sensor and a double intelligent algorithm of a simulated annealing algorithm and an extreme learning machine algorithm, which is characterized by comprising the following steps of:
step 1: the historical operation data collected by a central air-conditioning heating/refrigerating cycle field sensor is used as an initial data set Soff-lineAnd the selected sample comprises data of normal operation and data of fault operation;
step 2: training a central air-conditioning fault prediction and diagnosis classification neural network model of the extreme learning machine;
and step 3: judging whether the fault prediction and diagnosis classification neural network model meets the condition that the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum;
and 4, step 4: if the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, entering the step 7;
and 5: if the fault prediction and diagnosis misclassification rate of the central air conditioner is not met to be minimum, the step 6 is carried out; step 6: optimizing the central air conditioner fault prediction and diagnosis classification model coefficient of the extreme learning machine by adopting a simulated annealing algorithm to obtain a new central air conditioner fault prediction and diagnosis classification neural network model, and judging whether the central air conditioner fault prediction and diagnosis misclassification rate is minimum; if the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, entering the step 7; otherwise, continuing optimization until the minimum misclassification rate is met;
and 7: establishing a central air conditioner fault prediction and diagnosis classification model of a double intelligent algorithm;
and 8: a central air conditioner fault prediction and diagnosis classification model with a double intelligent algorithm is deployed on line;
and step 9: on-site central air conditioner operation real-time data set Son-lineWhether the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum or not is met;
step 10: if the on-site central air conditioner operates the real-time data set Son-lineIf the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, the step 12 is carried out;
step 11: if the on-site central air conditioner operates the real-time data set Son-lineIf the fault prediction and diagnosis misclassification rate of the central air conditioner is not met to be minimum, the field central air conditioner operation data set S is usedon-lineWith the initial data set Soff-lineMerging and continuing the step 2;
step 12: and outputting the central air conditioner fault prediction and diagnosis classification results.
The invention has the following specific advantages:
(1) the fault detection and diagnosis model takes the time domain signals collected by the central air conditioner field sensor as the characteristic vectors, overcomes the defect that the existing central air conditioner fault detection and diagnosis model established according to the laboratory equipment frequency domain signals cannot be deployed on the spot on line, has field feasibility and is convenient for engineering application.
(2) The single-layer feedforward neural network method for detecting and diagnosing the faults of the central air conditioner is built by utilizing the extreme learning machine, and is suitable for the mode recognition problem that the system has high-dimensional nonlinear characteristics and fewer fault data samples in the running process of the central air conditioner. Meanwhile, the extreme learning machine has good generalization capability and higher operation efficiency.
(3) The method is characterized in that the problem of improving the accuracy of central air conditioner fault mode identification is converted into an optimization problem with the minimum model misclassification rate, and the optimal solution with the minimum misclassification rate is obtained through a simulated annealing algorithm. The optimal solution found by the simulated annealing algorithm is independent of the selection of the initial solution and converges to the global optimal solution with a probability close to 1. Therefore, the problem of minimum misclassification rate of the central air conditioner is solved by adopting a simulated annealing algorithm, so that a global optimal solution is easier to obtain, and the accuracy and precision of fault detection and diagnosis model classification of the on-line deployed central air conditioner are improved.
Description of the drawings:
FIG. 1 is a flow chart of a central air conditioning fault prediction and diagnosis method based on a dual intelligent algorithm of a simulated annealing algorithm and an extreme learning machine algorithm according to the present invention;
FIG. 2 is a flow chart of a central air conditioning fault prediction and diagnosis method of the extreme learning machine of the present invention;
FIG. 3 is a flow chart for solving the minimum fault classification rate of central air conditioner fault prediction and diagnosis by using a simulated annealing algorithm.
The specific implementation mode is as follows:
a double intelligent algorithm central air conditioner fault prediction and diagnosis method comprises the following steps:
firstly, collecting a real-time data set S for the operation of a central air conditioner on siteon-line；
Secondly, the real-time data set S in the step of one is collectedon-lineInputting the data into a pre-established central air conditioner fault prediction and diagnosis classification neural network model based on a simulated annealing algorithm and an extreme learning machine algorithm to judge the real-time data set S in the step oneon-lineWhether the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum or not is met;
real-time data set S in the third and second stepson-lineAnd if the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, outputting a fault prediction and diagnosis classification result of the central air conditioner.
In the step two: if the real-time data set S in the step of "oneon-lineIf the fault prediction and diagnosis misclassification rate of the central air conditioner is not met to be minimum, the field central air conditioner operation data set S is usedon-lineAnd an initial data set S collected when a central air conditioner fault prediction and diagnosis classification model is establishedoff-lineAnd merging, and training a central air conditioner fault prediction and diagnosis classification neural network model. (this process isA data machine self-learning process is utilized. Because of the few types of fault data of the industrial system, new data sets need to be trained continuously, and fault data types need to be accumulated)
The method for constructing the central air-conditioning fault prediction and diagnosis classification neural network model comprises the following steps:
step 1: the historical operation data collected by the central air-conditioning heating/cooling heating and cooling cycle field sensor is used as an initial data set Soff-lineAnd the selected initial data set Soff-lineThe data comprises data with normal operation and data with failure operation;
step 2: training a central air-conditioning fault prediction and diagnosis classification neural network model of the extreme learning machine;
and step 3: judging whether the fault prediction and diagnosis classification neural network model meets the condition that the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum;
and 4, step 4: if the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, entering a step 7;
and 5: if the fault prediction and diagnosis misclassification rate of the central air conditioner is not met, entering a step 6; step 6: the method comprises the steps of optimizing central air conditioner fault prediction and diagnosis classification model coefficients of an extreme learning machine by adopting a simulated annealing algorithm (the simulated annealing algorithm is the existing method, but when the algorithm is applied to a central air conditioner fault prediction and diagnosis classification model, the design needs to be carried out by combining central air conditioner system parameters, the specific design method is claim 5, and a flow chart of the simulated annealing algorithm is shown in figure 3 of an attached drawing of the specification), obtaining a new central air conditioner fault prediction and diagnosis classification neural network model, and judging whether the new central air conditioner fault prediction and diagnosis classification neural network model meets the condition that the central air conditioner fault prediction and diagnosis misclassification rate is minimum; if the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, entering a step 7; otherwise, continuing optimization until the minimum misclassification rate is met;
and 7: establishing a central air conditioner fault prediction and diagnosis classification model of a double intelligent algorithm;
and 8: and deploying a central air conditioner fault prediction and diagnosis classification model of a double intelligent algorithm on line.
In step 2, the method for training the central air-conditioning fault prediction and diagnosis classification neural network model of the extreme learning machine specifically comprises the following steps:
step 1: the historical operation data collected by the central air-conditioning heating/cooling heating and cooling cycle field sensor is used as an initial data set Soff-lineAnd the selected sample comprises data of normal operation and data of fault operation;
step 2: constructing a single-layer feedforward neural network model consisting of 12 input neurons, 4 output neurons and L hidden layer neurons; wherein the input neurons are respectively at inhalation pressure x1Evaporator approach temperature x2The approach temperature x of the condenser3Return water temperature x of evaporator4Outlet water temperature x of evaporator5Condenser return water temperature x6Outlet water temperature x of condenser7Water supply pressure x of water separator8Water collector return pressure x9Current x10Voltage x11Heating/cooling and refrigerating capacity x of central air conditioner12(ii) a The output neurons are respectively of the normal operation type y of the central air conditioner1Evaporator failure type y2Type of condenser failure y3And compressor failure type y4；
And step 3: normalizing the 12 input neurons;
and 4, step 4: initial data set Soff-lineDivided into training sample sets StrainAnd a test sample set Stest(ii) a Wherein the training sample set StrainThe data comprises data of a normal operation type and data of an operation fault type, and the total number of the N discrete training samples
Figure BDA0002593168850000131
Wherein xi＝[xi1,xi2,…,xi12]T∈R12(ii) a Component xijA j-th input neuron representing an i-th sample, and i ═ 1,2, …, N; j is 1,2, …, 12. x is the number ofi11 st input neuron inspiratory pressure representing ith sample、xi2The 2 nd input neuron evaporator approach temperature, x, representing the ith samplei3The 3 rd input neuron condenser approach temperature, x, representing the ith samplei44 th input neuron evaporator Return Water temperature, x, representing the ith samplei5The evaporator leaving water temperature, x, of the 5 th input neuron representing the ith samplei6Condenser return water temperature, x, of the 6 th input neuron representing the ith samplei7Condenser outlet water temperature, x, of the 7 th input neuron representing the ith samplei8Water supply pressure, x, for the 8 th input neuron trap representing the ith samplei9The 9 th input neuron collector return water pressure, x, representing the ith samplei1010 th input neuron current, x, representing the ith samplei11The 11 th input neuron voltage, x, representing the ith samplei12The 12 th input neuron central air conditioning heating/cooling heating and cooling capacity representing the ith sample. x is the number ofiAs input to the single-layer feedforward neural network model. Etai＝[ηi1,ηi2,ηi3,ηi4]T∈R4Indicating normal operation result and faulty operation result data contained in the sample. Component ηi1The ith sample is the normal operation type of the central air conditioner, etai2Indicating the i-th sample as evaporator fault type, etai3Indicating the ith sample as the type of condenser fault, ηi4Indicating that the ith sample is a compressor failure type. Symbol [ 2 ]]TRepresents a transpose of a matrix;
and 5: selecting L as the number of hidden layer neurons;
step 6: adopting a random method to generate the weight value w of the kth neuron of the hidden layer and the 12 neurons of the input layerk＝[wk1,wk2,…,wk12]T(ii) a Generating a bias value b of the kth neuron of the hidden layer by adopting a random methodk，k＝1,2,…,L；
And 7: selecting an excitation function g (w) of a central air-conditioning fault prediction and classification neural network model of the extreme learning machinek·xi+bk) And i is 1,2, …, N, k is1,2,…,L；
And 8: using the weight value w calculated in step 61,...,wLAnd an offset value b1,...,bLAnd the excitation function in the step 7, and calculating the hidden layer output matrix H ═ H (w)1,...,wL,b1,...,bL,x1,...,xN)；
And step 9: calculating the weight beta from L neurons of the hidden layer to neurons of the output layer by using the hidden layer output matrix H obtained in the step 81,β2,…,βL]T(ii) a Wherein beta isk＝[βk1,βk2,βk3,βk4]TThe weight k representing the k-th neuron of the hidden layer to 4 neurons of the output layer is 1,2, …, L;
step 10: judging whether the beta in the step 9 meets the requirement of minimum output error of the single-layer feedforward neural network (the minimum output error is obtained by comparing all errors and then performing modular calculation, and the mathematical expression of the minimum output error is min | | yi- η i | |) and entering the step 11 if the minimum output error is obtained; otherwise, returning to the step 5;
step 11: outputting a central air conditioner fault prediction and classification model based on an extreme learning machine;
step 12: judging whether the central air-conditioning fault prediction and classification model based on the extreme learning machine in the step 11 meets the condition that the central air-conditioning fault prediction and diagnosis misclassification rate is minimum (the misclassification rate is minimum, namely in misclassification conversion, the value of the misclassification rate is enabled to be minimum
Figure BDA0002593168850000141
Minimum;
Figure BDA0002593168850000142
if yes, entering step 14, otherwise, directly entering step 13;
step 13: if the model does not meet the condition that the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, returning to the step 9, and optimizing the weight beta from L neurons of the hidden layer to neurons of the output layer by adopting a simulated annealing algorithm [ beta ]1,β2,…,βL]T；
Step 14: verifying the trained central air conditioner fault prediction and classification model by using the test sample;
step 15: judging whether the test sample meets the condition that the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum (the step is actually the test process of the step 14); if yes, go to step 16; otherwise, returning to the step 5;
step 16: and deploying a central air conditioner fault prediction and diagnosis classification model of a double intelligent algorithm on line.
Step 13, optimizing the weights beta from L neurons in the hidden layer to neurons in the output layer by adopting a simulated annealing algorithm1,β2,…,βL]TThe method comprises the following specific steps:
step 1: initializing relevant parameters of a simulated annealing algorithm, and selecting any initial solution beta in a feasible solution space based on a central air conditioner fault prediction and diagnosis model of an extreme learning machine;
step 2: calculating the misclassification rate W of the central air-conditioning fault prediction and diagnosis model generated by betaerror(β)；
And step 3: randomly generating beta*(New solution beta of feasible solution space based on extreme learning machine central air-conditioning fault prediction and diagnosis model*) Calculating by beta*Misclassification rate W of generated central air conditioner fault prediction and diagnosis modelerror(β*)；
And 4, step 4: judgment of Werror(β*)-Werror(β) is less than or equal to 0;
and 5: if Werror(β*)-WerrorIf beta is less than or equal to 0, then beta is used*Generating a central air conditioner fault prediction and diagnosis model, and entering the step 7;
step 6: if Werror(β*)-Werror(beta) > 0, then a bit at [0,1 ] is randomly generated]Random number in interval
Figure BDA0002593168850000151
Judgment of
Figure BDA0002593168850000152
Whether or not greater than
Figure BDA0002593168850000153
If it is
Figure BDA0002593168850000154
Then by beta*Generating a central air conditioner fault prediction and diagnosis classification model, and entering the step 7; if it is
Figure BDA0002593168850000155
The parameters of the central air conditioner fault prediction and diagnosis classification model are unchanged; the symbol exp () represents the exponent calculation;
and 7: judging whether the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum; if yes, entering step 8; otherwise, returning to the step 1, reducing the temperature control parameter t and continuing to operate;
and 8: deploying by beta online*And generating a central air conditioner fault prediction and diagnosis classification model with a double intelligent algorithm.
In step 2, the misclassification rate W of the central air conditioner fault prediction and diagnosis model generated by beta is calculatederrorThe method (β) is specifically as follows:
the quality evaluation of the central air-conditioning fault prediction and diagnosis classification model is evaluated by adopting fault misclassification rate conversion, and a misclassification rate calculation method is described by taking a training sample set as an example;
training sample set StrainThe data comprises normal operation data and operation fault data, and the total number of the N discrete training samples
Figure BDA0002593168850000161
W1The method is to classify the sample of the central air conditioner fault prediction and diagnosis model output as the normal operation type as the fault type (meaning that the sample output as the normal operation type is classified as the fault type, namely a normal operation signal, but because of uncertain factors such as model error or sensor fault, the signal of the correct operation is identified as the fault type by mistake when the fault prediction and diagnosis is carried outA signal of operation; or misidentify a malfunctioning signal as a properly functioning signal. Such misclassifications have a particularly large impact on the model. Illustrating that normal samples of the evaporator are misclassified into a sample set with evaporator failure; or misclassifying the evaporator fault sample into a normal evaporator sample set, and the like), or misclassifying the sample with the model output as the fault type into a normal operation type set;
W2the method comprises the steps of classifying a sample which is output as a fault state by a model into a fault type, but not belonging to a data set of a correct fault type sample; if the original output type is evaporator fault but misclassification is condenser fault or compressor fault, the original output type is condenser fault but misclassification is evaporator fault or compressor fault, and the original output type is compressor fault but misclassification is evaporator fault or condenser fault;
W3is a set of classification rates for which the output type is correct, and satisfies
Figure BDA0002593168850000162
Wherein the symbols
Figure BDA0002593168850000163
Is a direct sum operation symbol;
the minimum misclassification transformation of the central air-conditioning fault prediction and diagnosis is as follows:
Figure BDA0002593168850000164
Figure BDA0002593168850000165
in which ξij(beta) represents a central air conditioner fault prediction and diagnosis classification model classification function by a dual intelligent algorithm; y isiRepresenting output layer neurons trained by neural networks, yi＝[yi1,yi2,yi3,yi4]TI is 1,2, …, N, and satisfies the output errorMinimum min yi-ηi||。
In step 3, the 12 input neuron parameters are normalized to 0-1 numerical values, and the specific normalization method is as follows:
(1) inspiratory pressure parameter x1Is normalized to obtain
Figure BDA0002593168850000171
Figure BDA0002593168850000172
Wherein
Figure BDA0002593168850000173
The air suction pressure under the working condition is designed for the central air-conditioning system,
Figure BDA0002593168850000174
is a normalized inspiratory pressure parameter;
(2) evaporator approach temperature parameter x2Is normalized to obtain
Figure BDA0002593168850000175
Figure BDA0002593168850000176
Wherein
Figure BDA0002593168850000177
The evaporator approaches the temperature under the design working condition of the central air-conditioning system,
Figure BDA0002593168850000178
the evaporator approach temperature parameter after normalization;
(3) condenser approach temperature x3Is normalized to obtain
Figure BDA0002593168850000179
Figure BDA00025931688500001710
Wherein
Figure BDA00025931688500001711
The approach temperature of the condenser under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500001712
the normalized condenser approach temperature parameter is obtained;
(4) return water temperature x of evaporator4Is normalized to obtain
Figure BDA00025931688500001713
Figure BDA00025931688500001714
Wherein
Figure BDA00025931688500001715
The return water temperature of the evaporator under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500001716
the normalized return water temperature parameter of the evaporator is obtained;
(5) water outlet temperature x of evaporator5Is normalized to obtain
Figure BDA00025931688500001717
Figure BDA00025931688500001718
Wherein
Figure BDA00025931688500001719
The outlet water temperature of the evaporator under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500001720
the normalized outlet water temperature parameter of the evaporator is obtained;
(6) condenser return water temperature x6Is normalized to obtain
Figure BDA00025931688500001721
Figure BDA00025931688500001722
Wherein
Figure BDA00025931688500001723
The return water temperature of the condenser under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500001724
the normalized condenser return water temperature parameter is obtained;
(7) condenser outlet water temperature x7Is normalized to obtain
Figure BDA00025931688500001725
Figure BDA00025931688500001726
Wherein
Figure BDA00025931688500001727
Designing the outlet water temperature of the condenser under the working condition for the central air-conditioning system;
Figure BDA00025931688500001728
the water outlet temperature parameter is the normalized condenser water outlet temperature parameter;
(8) water supply pressure x of water separator8Is normalized to obtain
Figure BDA00025931688500001729
Figure BDA00025931688500001730
Wherein
Figure BDA00025931688500001731
The water supply pressure of the water distributor under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500001732
the water supply pressure parameter of the water separator after normalization;
(9) water collector return pressure x9Normalized to obtainTo
Figure BDA00025931688500001733
Figure BDA00025931688500001734
Wherein
Figure BDA00025931688500001735
The backwater pressure of the water collector under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500001736
the water pressure parameter is the normalized water collector return water pressure parameter;
(10) current x10Is normalized to obtain
Figure BDA00025931688500001737
Figure BDA00025931688500001738
Wherein
Figure BDA00025931688500001739
Rated current for safe operation of the central air-conditioning system;
Figure BDA00025931688500001740
is a normalized current parameter;
(11) voltage x11Is normalized to obtain
Figure BDA0002593168850000181
Figure BDA0002593168850000182
Wherein
Figure BDA0002593168850000183
Rated voltage for safe operation of the central air-conditioning system;
Figure BDA0002593168850000184
the normalized voltage parameter is obtained;
(13) heating and cooling capacity x of central air conditioner12Is normalized to obtain
Figure BDA0002593168850000185
Figure BDA0002593168850000186
Wherein
Figure BDA0002593168850000187
The calculation method is as follows for the nominal heating capacity of the central air conditioner:
Figure BDA0002593168850000188
wherein K1Correcting coefficient of heating capacity or cooling capacity of the central air conditioning unit; k2A defrosting correction coefficient of the central air conditioning unit; k, the comprehensive heat transfer coefficient of the building; t isnHeating and cooling indoor design temperature, TpCalculating the outdoor temperature in the heating season or the refrigerating season;
Figure BDA0002593168850000189
heating and refrigerating capacity parameters of the normalized central air conditioner;
where the symbol | | | represents the modulus of the vector.
Step 8 calculates the hidden layer output matrix H ═ H (w)1,...,wL,b1,...,bL,x1,...,xN)：
Figure BDA00025931688500001810
Wherein x1＝[x11,x12,…,x112]TRepresents the 1 st sample parameter in the training dataset: the air-conditioning system comprises an air suction pressure, an evaporator approach temperature, a condenser approach temperature, an evaporator return water temperature, an evaporator outlet water temperature, a condenser return water temperature, a condenser outlet water temperature, a water distributor supply pressure, a water collector return water pressure, current, voltage and central air-conditioning heating and refrigerating capacity; these 12 parameters are used as the middle of the extreme learning machine algorithmThe method comprises the steps of (1) inputting layer parameters of a central air conditioner fault prediction and diagnosis neural network model;
wherein xN＝[xN1,xN2,…,xN12]TRepresents the nth sample parameter in the training dataset: the air-conditioning system comprises an air suction pressure, an evaporator approach temperature, a condenser approach temperature, an evaporator return water temperature, an evaporator outlet water temperature, a condenser return water temperature, a condenser outlet water temperature, a water distributor supply pressure, a water collector return water pressure, current, voltage and central air-conditioning heating and refrigerating capacity; the 12 parameters are used as input layer parameters of a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
wherein w1＝[w11,w12,…,w112]Representing the weight values of the 1 st neuron of the hidden layer and the 12 neurons of the input layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm, and the component w11Representing the weight value of the 1 st neuron of the hidden layer and the inspiration pressure neuron of the input layer; w is a12Representing the weight value of the 1 st neuron of the hidden layer and the approximate temperature neuron of the evaporator of the input layer; w is a13Representing the weight value of the 1 st neuron of the hidden layer and the input layer condenser approaching temperature neuron; w is a14Representing the weight values of the 1 st neuron of the hidden layer and the return water temperature neuron of the evaporator of the input layer; w is a15Representing the weight values of the 1 st neuron of the hidden layer and the water outlet temperature neuron of the input layer evaporator; w is a16Representing the weight values of the 1 st neuron of the hidden layer and the return water temperature neuron of the condenser of the input layer; w is a17Representing the weight values of the 1 st neuron of the hidden layer and the water outlet temperature neuron of the input layer condenser; w is a18Representing the weight values of the 1 st neuron of the hidden layer and the water supply pressure neuron of the input layer water separator; w is a19Representing the weight value of the 1 st neuron of the hidden layer and the backwater pressure neuron of the water collector of the input layer; w is a110Representing the weight value of the 1 st neuron of the hidden layer and the current neuron of the input layer; w is a111Representing the weight values of the 1 st neuron of the hidden layer and the input layer voltage neuron; w is a112Central air-conditioning heating of 1 st neuron of representing hidden layer and input layerAnd the weight value of the refrigerating output neuron;
wherein wL＝[wL1,wL2,…,wL12]Representing the weight values of the Lth neuron of the hidden layer and the 12 neurons of the input layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm, and the component wL1Representing the weight value of the Lth neuron of the hidden layer and the inspiration pressure neuron of the input layer; w is aL2The weight value of the Lth neuron of the hidden layer and the approximate temperature neuron of the evaporator of the input layer is represented; w is aL3Representing the weight value of the Lth neuron of the hidden layer and the input layer condenser approaching temperature neuron; w is aL4Representing the weight values of the Lth neuron of the hidden layer and the return water temperature neuron of the evaporator of the input layer; w is aL5Representing the weight values of the L-th neuron of the hidden layer and the water outlet temperature neuron of the input layer evaporator; w is aL6Representing the weight values of the Lth neuron of the hidden layer and the return water temperature neuron of the condenser of the input layer; w is aL7Representing the weight values of the Lth neuron of the hidden layer and the water outlet temperature neuron of the input layer condenser; w is aL8Representing the weight value of the L-th neuron of the hidden layer and the water supply pressure neuron of the input layer water separator; w is aL9Representing the weight value of the Lth neuron of the hidden layer and the backwater pressure neuron of the water collector of the input layer; w is aL10Representing the weight value of the 1 st neuron of the hidden layer and the current neuron of the input layer; w is aL11Representing the weight values of the Lth neuron of the hidden layer and the voltage neuron of the input layer; w is aL12And the weight values of the Lth neuron of the hidden layer and the heating and cooling capacity neuron of the central air conditioner of the input layer are represented. "·" denotes an inner product operation; with wL·x1The description is given for the sake of example: w is aL·x1＝wL1×x11+wL2×x12+…wL12×x112∈R。
Excitation function g (w)k·xi+bk) The central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm comprises the suction pressure, the evaporator approach temperature, the condenser approach temperature, the evaporator return water temperature, the evaporator outlet water temperature and the condensation of 12 input layer neuronsAnd mapping linear spaces of the water returning temperature of the device, the water outlet temperature of the condenser, the water supply pressure of the water distributor, the water returning pressure of the water collector, the current, the voltage and the heating and refrigerating capacity of the central air conditioner into a linear space containing L hidden layer characteristics in the extreme learning machine. The specific selection method of the excitation function is as follows:
trigonometric function: g (w)k·xi+bk)＝cos(wk·xi+bk)；
Gaussian function:
Figure BDA0002593168850000191
sigmoid function:
Figure BDA0002593168850000192
hyperbolic sine function:
Figure BDA0002593168850000201
radial basis function: g (wk. xi + b)k)＝exp(-bk·xi-wk||)。
And step 9: calculating the weight hidden layer output weight beta from the hidden layer L neurons to the output layer neurons according to beta H + Y1,β2,…,βL]；
Wherein beta is1＝[β11,β12,β13,β14]TMiddle component beta11The central air conditioner fault prediction and diagnosis neural network model based on the extreme learning machine algorithm is characterized by comprising 1 st neuron and 1 st output layer neuron of a hidden layer, wherein the 1 st neuron and the 1 st output layer neuron are respectively connected with a central air conditioner; component beta12Representing the weight values of the evaporator fault types of the 1 st neuron and the 2 nd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta13Representing the weight values of the condenser fault types of the 1 st neuron and the 3 rd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta14Representing the weight values of compressor fault types of 1 st neuron and 4 th output layer neuron of a hidden layer in a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
wherein beta is2＝[β21,β22,β23,β24]TMiddle component beta21The central air conditioner fault prediction and diagnosis neural network model based on the extreme learning machine algorithm is characterized by comprising the following steps of (1) representing the weight value of the normal operation type of the central air conditioner of the hidden layer 2 th neuron and the hidden layer 1 st output layer neuron; component beta22Representing the weight values of the evaporator fault types of the neurons of the No. 2 and the No. 2 output layers of the hidden layers in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta23Representing the weight values of the condenser fault types of the 2 nd neuron and the 3 rd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta24Representing the weight values of compressor fault types of a hidden layer 2 nd neuron and a hidden layer 4 th output layer neuron in a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
wherein beta isL＝[βL1,βL2,βL3,βL4]TMiddle component beta11The central air conditioner fault prediction and diagnosis neural network model based on the extreme learning machine algorithm is characterized by comprising the following steps of (1) representing the weight value of the normal operation type of the central air conditioner of the hidden layer Lth neuron and the 1 st output layer neuron; component beta12Representing the weight values of the evaporator fault types of the L-th neuron and the 2 nd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta13Representing the weight values of the condenser fault types of the Lth neuron and the 3 rd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta14Representing compressor failure of Lth neuron and 4 th output layer neuron of hidden layer in central air-conditioning fault prediction and diagnosis neural network model based on extreme learning machine algorithmA weight value of the barrier type;
H+output matrix H ═ H (w) for hidden layers1,...,wL,b1,...,bL,x1,...,xN) Moore-Penrose (Moore-Penrose) generalized inverse matrix of (g); when H is presentTH is not singular: h + - (H)TH)-1HTOr H + ═ HT(HTH)-1(ii) a Wherein ()-1Represents the inverse of the matrix ()TRepresents a transpose of a matrix;
Y＝[y1,y2,…,yN]Tzero-error approximation of N discrete training samples for training
Figure BDA0002593168850000211
Of the neural network, wherein yi＝[yi1,yi2,yi3,yi4]T，i＝1,2,…,N。
At y1Middle, component y11Representing that the 1 st neuron output after the training of the neural network model of the extreme learning machine is carried out on the 1 st sample by zero error approximation is the normal operation type of the central air conditioner; component y12Representing that the 2 nd neuron output after the 1 st training sample is subjected to extreme learning machine neural network model training is an evaporator fault type by zero error approximation; component y13Representing that the 3 rd neuron output after the training of the neural network model of the extreme learning machine is carried out on the 1 st training sample by using zero error approximation is a condenser fault type; component y14And representing that the 4 th neuron output after the extreme learning machine neural network model training is carried out on the 1 st training sample by zero error approximation is a compressor fault type.
At y2Middle, component y21Representing that the 1 st neuron output after the 2 nd sample is subjected to extreme learning machine neural network model training is a normal operation type of the central air conditioner by using zero error approximation; component y22Representing that the 2 nd neuron output after the training of the neural network model of the extreme learning machine is carried out on the 2 nd training sample by zero error approximation is an evaporator fault type; component y23Representing that the 3 rd neuron output after the 2 nd training sample is subjected to extreme learning machine neural network model training is a condenser fault type by zero error approximation; component y24And representing that the 4 th neuron output after the extreme learning machine neural network model training is carried out on the 2 nd training sample by zero error approximation is a compressor fault type.
At yNMiddle, component yN1Representing that the 1 st neuron output after the training of the neural network model of the extreme learning machine is carried out on the Nth training sample by using zero error approximation is the normal operation type of the central air conditioner; component yN2Representing that the Nth neuron output after the Nth training sample is subjected to extreme learning machine neural network model training by using zero error approximation is an evaporator fault type; component yN3Representing that the 3 rd neuron output after the training of the neural network model of the extreme learning machine is carried out on the Nth training sample by using zero error approximation is a condenser fault type; component yN4And representing that the 4 th neuron output after the training of the extreme learning machine neural network model is carried out on the Nth training sample by using zero error approximation is a compressor fault type.
yiThe calculation method is
Figure BDA0002593168850000212
And the minimum min y of the error output by the neural network model is satisfiedi-ηi||，i＝1,2,…,N，k＝1,2,…,L。
The system comprises a data acquisition module, a data judgment module and a fault prediction and diagnosis type data output module:
the data acquisition module acquires a real-time data set S for the operation of the central air conditioner on siteon-line；
A data judgment module for collecting the real-time data set S in the data acquisition moduleon-lineInputting the data into a pre-established central air conditioner fault prediction and diagnosis classification neural network model based on a simulated annealing algorithm and an extreme learning machine algorithm to judge whether the real-time data set in the step one meets the central air conditioner faultThe prediction and diagnosis misclassification rate is minimum;
the fault prediction and diagnosis type data output module: real-time data set S in data judgment moduleon-lineAnd outputting the central air conditioner fault prediction and diagnosis classification result if the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum.
The present invention will be described in detail below with reference to the accompanying drawings.
Step 1: the historical operation data collected by a central air-conditioning heating/refrigerating cycle field sensor is used as an initial data set Soff-lineAnd the selected sample comprises data of normal operation and data of fault operation;
step 2: training a central air-conditioning fault prediction and diagnosis classification neural network model of the extreme learning machine;
and step 3: judging whether the fault prediction and diagnosis classification neural network model meets the condition that the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum;
and 4, step 4: if the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, entering the step 7;
and 5: if the fault prediction and diagnosis misclassification rate of the central air conditioner is not met to be minimum, the step 6 is carried out;
step 6: optimizing the central air conditioner fault prediction and diagnosis classification model coefficient of the extreme learning machine by adopting a simulated annealing algorithm to obtain a new central air conditioner fault prediction and diagnosis classification neural network model, and judging whether the central air conditioner fault prediction and diagnosis misclassification rate is minimum; if the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, entering the step 7; otherwise, continuing optimization until the minimum misclassification rate is met;
and 7: establishing a central air conditioner fault prediction and diagnosis classification model of a double intelligent algorithm;
and 8: a central air conditioner fault prediction and diagnosis classification model with a double intelligent algorithm is deployed on line;
and step 9: on-site central air conditioner operation real-time data set Son-lineWhether the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum or not is met;
step 10: if the on-site central air conditioner operates the real-time data set Son-lineIf the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, the step 12 is carried out;
step 11: if the on-site central air conditioner operates the real-time data set Son-lineIf the fault prediction and diagnosis misclassification rate of the central air conditioner is not met to be minimum, the field central air conditioner operation data set S is usedon-lineWith the initial data set Soff-lineMerging and continuing the step 2;
step 12: and outputting the central air conditioner fault prediction and diagnosis classification results.
FIG. 2 is a flow chart of a central air conditioning fault prediction and diagnosis method of the extreme learning machine of the present invention; the calculation process as shown is as follows:
step 1: the historical operation data collected by a central air-conditioning heating/refrigerating cycle field sensor is used as an initial data set Soff-lineAnd the selected sample comprises data of normal operation and data of fault operation;
step 2: constructing a single-layer feedforward neural network model consisting of 12 input neurons, 4 output neurons and L hidden layer neurons; wherein the input neurons are respectively at inhalation pressure x1Evaporator approach temperature x2The approach temperature x of the condenser3Return water temperature x of evaporator4Outlet water temperature x of evaporator5Condenser return water temperature x6Outlet water temperature x of condenser7Water supply pressure x of water separator8Water collector return pressure x9Current x10Voltage x11Heating/cooling capacity x of central air conditioner12(ii) a The output neurons are respectively of the normal operation type y of the central air conditioner1Evaporator failure type y2Type of condenser failure y3Compressor failure type y4；
In step 3, the 12 input neuron parameters are normalized to 0-1 numerical values, and the specific normalization method is as follows:
(1) inspiratory pressure parameter x1Is normalized to obtain
Figure BDA0002593168850000231
Figure BDA0002593168850000232
Wherein
Figure BDA0002593168850000233
The air suction pressure under the working condition is designed for the central air-conditioning system,
Figure BDA0002593168850000234
is a normalized inspiratory pressure parameter;
(2) evaporator approach temperature parameter x2Is normalized to obtain
Figure BDA0002593168850000235
Figure BDA0002593168850000236
Wherein
Figure BDA0002593168850000237
The evaporator approaches the temperature under the design working condition of the central air-conditioning system,
Figure BDA0002593168850000238
the evaporator approach temperature parameter after normalization;
(3) condenser approach temperature x3Is normalized to obtain
Figure BDA0002593168850000239
Figure BDA00025931688500002310
Wherein
Figure BDA00025931688500002311
The approach temperature of the condenser under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500002312
the normalized condenser approach temperature parameter is obtained;
(4) return water temperature x of evaporator4Is normalized to obtain
Figure BDA00025931688500002313
Figure BDA00025931688500002314
Wherein
Figure BDA00025931688500002315
The return water temperature of the evaporator under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500002316
the normalized return water temperature parameter of the evaporator is obtained;
(5) water outlet temperature x of evaporator5Is normalized to obtain
Figure BDA00025931688500002317
Figure BDA00025931688500002318
Wherein
Figure BDA00025931688500002319
The outlet water temperature of the evaporator under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500002320
the normalized outlet water temperature parameter of the evaporator is obtained;
(6) condenser return water temperature x6Is normalized to obtain
Figure BDA00025931688500002321
Figure BDA00025931688500002322
Wherein
Figure BDA00025931688500002323
The return water temperature of the condenser under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500002324
the normalized condenser return water temperature parameter is obtained;
(7) condenser outlet water temperature x7Is normalized to obtain
Figure BDA00025931688500002325
Figure BDA00025931688500002326
Wherein
Figure BDA00025931688500002327
Designing the outlet water temperature of the condenser under the working condition for the central air-conditioning system;
Figure BDA00025931688500002328
the water outlet temperature parameter is the normalized condenser water outlet temperature parameter;
(8) water supply pressure x of water separator8Is normalized to obtain
Figure BDA00025931688500002329
Figure BDA00025931688500002330
Wherein
Figure BDA00025931688500002331
The water supply pressure of the water distributor under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500002332
the water supply pressure parameter of the water separator after normalization;
(9) water collector return pressure x9Is normalized to obtain
Figure BDA00025931688500002333
Figure BDA00025931688500002334
Wherein
Figure BDA00025931688500002335
The backwater pressure of the water collector under the working condition is designed for the central air-conditioning system;
Figure BDA00025931688500002336
the water pressure parameter is the normalized water collector return water pressure parameter;
(10) current x10Is normalized to obtain
Figure BDA0002593168850000241
Figure BDA0002593168850000242
Wherein
Figure BDA0002593168850000243
Rated current for safe operation of the central air-conditioning system;
Figure BDA0002593168850000244
is a normalized current parameter;
(11) voltage x11Is normalized to obtain
Figure BDA0002593168850000245
Figure BDA0002593168850000246
Wherein
Figure BDA0002593168850000247
Rated voltage for safe operation of the central air-conditioning system;
Figure BDA0002593168850000248
the normalized voltage parameter is obtained;
(12) heating/cooling capacity x of central air conditioner12Is normalized to obtain
Figure BDA0002593168850000249
Figure BDA00025931688500002410
Wherein
Figure BDA00025931688500002411
The calculation method is as follows for the nominal heating capacity of the central air conditioner:
Figure BDA00025931688500002412
wherein K1Correcting coefficient of heating capacity or cooling capacity of the central air conditioning unit; k2A defrosting correction coefficient of the central air conditioning unit; k, the comprehensive heat transfer coefficient of the building; t isnHeating/cooling indoor design temperature, TpCalculating the outdoor temperature in the heating season or the refrigerating season;
Figure BDA00025931688500002413
the normalized central air-conditioning heating/cooling capacity parameter is obtained;
where the symbol | | | represents the modulus of the vector.
And 4, step 4: initial data set Soff-lineDivided into training sample sets StrainAnd a test sample set Stest(ii) a Wherein the training sample set StrainThe data comprises data of a normal operation type and data of an operation fault type, and the total number of the N discrete training samples
Figure BDA00025931688500002414
Wherein xi＝[xi1,xi2,…,xi12]T∈R12. Component xijA j-th input neuron representing an i-th sample, and i ═ 1,2, …, N; j is 1,2, …, 12. x is the number ofi11 st input neuron inspiratory pressure, x, representing the ith samplei2The 2 nd input neuron evaporator approach temperature, x, representing the ith samplei3The 3 rd input neuron condenser approach temperature, x, representing the ith samplei44 th input neuron evaporator Return Water temperature, x, representing the ith samplei5The evaporator leaving water temperature, x, of the 5 th input neuron representing the ith samplei6Condenser return water temperature, x, of the 6 th input neuron representing the ith samplei77 th input god representing ith sampleOutlet water temperature x of the first condenseri8Water supply pressure, x, for the 8 th input neuron trap representing the ith samplei9The 9 th input neuron collector return water pressure, x, representing the ith samplei1010 th input neuron current, x, representing the ith samplei11The 11 th input neuron voltage, x, representing the ith samplei12And (3) representing the heating/cooling capacity of the central air conditioner of the 12 th input neuron of the ith sample. x is the number ofiAs input to the single-layer feedforward neural network model. Etai＝[ηi1,ηi2,ηi3,ηi4]T∈R4Indicating normal operation result and faulty operation result data contained in the sample. Component ηi1The ith sample is the normal operation type of the central air conditioner, etai2Indicating the i-th sample as evaporator fault type, etai3Indicating the ith sample as the type of condenser fault, ηi4Indicating that the ith sample is a compressor failure type. Symbol [ 2 ]]TRepresents a transpose of a matrix;
and 5: selecting L as the number of hidden layer neurons;
step 6: adopting a random method to generate the weight value w of the kth neuron of the hidden layer and the 12 neurons of the input layerk＝[wk1,wk2,…,wk12]T(ii) a Generating a bias value b of the kth neuron of the hidden layer by adopting a random methodk，k＝1,2,…,L；
And 7: selecting an excitation function g (w) of a central air-conditioning fault prediction and classification neural network model of the extreme learning machinek·xi+bk) And i is 1,2, …, N, k is 1,2, …, L;
and 8: using the weight value w calculated in step 61,…,wLAnd an offset value b1,...,bLAnd the excitation function in the step 7, and calculating the hidden layer output matrix H ═ H (w)1,...,wL,b1,...,bL,x1,...,xN)：
Figure BDA0002593168850000251
Wherein x1＝[x11,x12,…,x112]TRepresents the 1 st sample parameter in the training dataset: the air-conditioning system comprises an air suction pressure, an evaporator approach temperature, a condenser approach temperature, an evaporator return water temperature, an evaporator outlet water temperature, a condenser return water temperature, a condenser outlet water temperature, a water distributor supply water pressure, a water collector return water pressure, current, voltage and central air-conditioning heating/refrigerating capacity; the 12 parameters are used as input layer parameters of a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
wherein xN＝[xN1,xN2,…,xN12]TRepresents the nth sample parameter in the training dataset: the air-conditioning system comprises an air suction pressure, an evaporator approach temperature, a condenser approach temperature, an evaporator return water temperature, an evaporator outlet water temperature, a condenser return water temperature, a condenser outlet water temperature, a water distributor supply water pressure, a water collector return water pressure, current, voltage and central air-conditioning heating/refrigerating capacity; the 12 parameters are used as input layer parameters of a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
wherein w1＝[w11,w12,…,w112]Representing the weight values of the 1 st neuron of the hidden layer and the 12 neurons of the input layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm, and the component w11Representing the weight value of the 1 st neuron of the hidden layer and the inspiration pressure neuron of the input layer; w is a12Representing the weight value of the 1 st neuron of the hidden layer and the approximate temperature neuron of the evaporator of the input layer; w is a13Representing the weight value of the 1 st neuron of the hidden layer and the input layer condenser approaching temperature neuron; w is a14Representing the weight values of the 1 st neuron of the hidden layer and the return water temperature neuron of the evaporator of the input layer; w is a15Representing the weight values of the 1 st neuron of the hidden layer and the water outlet temperature neuron of the input layer evaporator; w is a16Representing the weight values of the 1 st neuron of the hidden layer and the return water temperature neuron of the condenser of the input layer; w is a17Indicating the 1 st of the hidden layerThe weight value of the neuron and the input layer condenser outlet water temperature neuron; w is a18Representing the weight values of the 1 st neuron of the hidden layer and the water supply pressure neuron of the input layer water separator; w is a19Representing the weight value of the 1 st neuron of the hidden layer and the backwater pressure neuron of the water collector of the input layer; w is a110Representing the weight value of the 1 st neuron of the hidden layer and the current neuron of the input layer; w is a111Representing the weight values of the 1 st neuron of the hidden layer and the input layer voltage neuron; w is a112Representing the weight values of the 1 st neuron of the hidden layer and the heating/cooling capacity neuron of the central air conditioner of the input layer;
wherein wL＝[wL1,wL2,…,wL12]Representing the weight values of the Lth neuron of the hidden layer and the 12 neurons of the input layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm, and the component wL1Representing the weight value of the Lth neuron of the hidden layer and the inspiration pressure neuron of the input layer; w is aL2The weight value of the Lth neuron of the hidden layer and the approximate temperature neuron of the evaporator of the input layer is represented; w is aL3Representing the weight value of the Lth neuron of the hidden layer and the input layer condenser approaching temperature neuron; w is aL4Representing the weight values of the Lth neuron of the hidden layer and the return water temperature neuron of the evaporator of the input layer; w is aL5Representing the weight values of the L-th neuron of the hidden layer and the water outlet temperature neuron of the input layer evaporator; w is aL6Representing the weight values of the Lth neuron of the hidden layer and the return water temperature neuron of the condenser of the input layer; w is aL7Representing the weight values of the Lth neuron of the hidden layer and the water outlet temperature neuron of the input layer condenser; w is aL8Representing the weight value of the L-th neuron of the hidden layer and the water supply pressure neuron of the input layer water separator; w is aL9Representing the weight value of the Lth neuron of the hidden layer and the backwater pressure neuron of the water collector of the input layer; w is aL10Representing the weight value of the 1 st neuron of the hidden layer and the current neuron of the input layer; w is aL11Representing the weight values of the Lth neuron of the hidden layer and the voltage neuron of the input layer; w is aL12And the weight values of the Lth neuron of the hidden layer and the heating/cooling capacity neuron of the central air conditioner of the input layer are represented.
"·" denotes an inner product operation; with wL·x1The description is given for the sake of example: w is aL·x1＝wL1×x11+wL2×x12+…wL12×x112∈R。
Excitation function g (w)k·xi+bk) A central air conditioner fault prediction and diagnosis neural network model based on an extreme learning machine algorithm maps a linear space containing 12 input layer neuron suction pressures, evaporator approach temperatures, condenser approach temperatures, evaporator return water temperatures, evaporator outlet water temperatures, condenser return water temperatures, condenser outlet water temperatures, water distributor supply pressures, water collector return water pressures, currents, voltages and central air conditioner heating/cooling capacities into a linear space containing L hidden layer characteristics in an extreme learning machine. The excitation function may be selected from any of the following functions:
trigonometric function: g (w)k·xi+bk)＝cos(wk·xi+bk)；
Gaussian function:
Figure BDA0002593168850000261
sigmoid function:
Figure BDA0002593168850000262
hyperbolic sine function:
Figure BDA0002593168850000263
radial basis function: g (w)k·xi+bk)＝exp(-bk·||xi-wk||)。
And step 9: using the hidden layer output matrix H obtained in step 8, according to β ═ H+Y calculates the weight from L neurons in the hidden layer to the output layer neuron, and the hidden layer output weight beta is [ beta ]1,β2,…,βL]；
Wherein beta is1＝[β11,β12,β13,β14]TMiddle component beta11The central air conditioner fault prediction and diagnosis neural network model based on the extreme learning machine algorithm is characterized by comprising 1 st neuron and 1 st output layer neuron of a hidden layer, wherein the 1 st neuron and the 1 st output layer neuron are respectively connected with a central air conditioner; component beta12Representing the weight values of the evaporator fault types of the 1 st neuron and the 2 nd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta13Representing the weight values of the condenser fault types of the 1 st neuron and the 3 rd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta14Representing the weight values of compressor fault types of 1 st neuron and 4 th output layer neuron of a hidden layer in a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
wherein beta is2＝[β21,β22,β23,β24]TMiddle component beta21The central air conditioner fault prediction and diagnosis neural network model based on the extreme learning machine algorithm is characterized by comprising the following steps of (1) representing the weight value of the normal operation type of the central air conditioner of the hidden layer 2 th neuron and the hidden layer 1 st output layer neuron; component beta22Representing the weight values of the evaporator fault types of the neurons of the No. 2 and the No. 2 output layers of the hidden layers in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta23Representing the weight values of the condenser fault types of the 2 nd neuron and the 3 rd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta24Representing the weight values of compressor fault types of a hidden layer 2 nd neuron and a hidden layer 4 th output layer neuron in a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
wherein beta isL＝[βL1,βL2,βL3,βL4]TMiddle component beta11Representing L < th > hidden layer in central air-conditioning fault prediction and diagnosis neural network model based on extreme learning machine algorithmThe weight value of the normal operation type of the central air conditioner of each neuron and the 1 st output layer neuron; component beta12Representing the weight values of the evaporator fault types of the L-th neuron and the 2 nd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta13Representing the weight values of the condenser fault types of the Lth neuron and the 3 rd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta14Representing the weight values of compressor fault types of the Lth neuron and the 4 th output layer neuron of a hidden layer in a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
H+output matrix H ═ H (w) for hidden layers1,...,wL,b1,...,bL,x1,...,xN) Moore-Penrose (Moore-Penrose) generalized inverse matrix of (g); h+The following two methods can be used:
when H is presentTH is not singular: h+＝(HTH)-1HTOr H+＝HT(HTH)-1(ii) a Wherein ()-1Represents the inverse of the matrix ()TRepresenting the transpose of the matrix.
Solving by ridge regression algorithm, i.e. H+＝(HTH+λE)-1HTOr H+＝HT(HTH+λE)-1(ii) a Where λ represents a very small positive number, E represents HTH is the same as the unit matrix.
Y＝[y1,y2,…,yN]TZero-error approximation of N discrete training samples for training
Figure BDA0002593168850000281
Of the neural network, wherein yi＝[yi1,yi2,yi3,yi4]T，i＝1,2,…,N。
At y1Middle, component y11Representation from zero error approximationThe 1 st neuron output after the 1 st sample is subjected to extreme learning machine neural network model training is a normal operation type of the central air conditioner; component y12Representing that the 2 nd neuron output after the 1 st training sample is subjected to extreme learning machine neural network model training is an evaporator fault type by zero error approximation; component y13Representing that the 3 rd neuron output after the training of the neural network model of the extreme learning machine is carried out on the 1 st training sample by using zero error approximation is a condenser fault type; component y14And representing that the 4 th neuron output after the extreme learning machine neural network model training is carried out on the 1 st training sample by zero error approximation is a compressor fault type.
At y2Middle, component y21Representing that the 1 st neuron output after the 2 nd sample is subjected to extreme learning machine neural network model training is a normal operation type of the central air conditioner by using zero error approximation; component y22Representing that the 2 nd neuron output after the training of the neural network model of the extreme learning machine is carried out on the 2 nd training sample by zero error approximation is an evaporator fault type; component y23Representing that the 3 rd neuron output after the 2 nd training sample is subjected to extreme learning machine neural network model training is a condenser fault type by zero error approximation; component y24And representing that the 4 th neuron output after the extreme learning machine neural network model training is carried out on the 2 nd training sample by zero error approximation is a compressor fault type.
At yNMiddle, component yN1Representing that the 1 st neuron output after the training of the neural network model of the extreme learning machine is carried out on the Nth training sample by using zero error approximation is the normal operation type of the central air conditioner; component yN2Representing that the Nth neuron output after the Nth training sample is subjected to extreme learning machine neural network model training by using zero error approximation is an evaporator fault type; component yN3Representing that the 3 rd neuron output after the training of the neural network model of the extreme learning machine is carried out on the Nth training sample by using zero error approximation is a condenser fault type; component yN4Representing extreme learning machine with zero-error approximation from the Nth training sampleAnd 4, the 4 th neuron output after the neural network model is trained is a compressor fault type.
yiThe calculation method is
Figure BDA0002593168850000282
Step 10: judging whether beta is satisfied
Figure BDA0002593168850000283
I.e. the output error y of the single-layer neural networki-ηi| | minimum? If yes, entering step 11; otherwise, returning to the step 5;
step 11: outputting a central air conditioner fault prediction and classification model based on an extreme learning machine;
step 12: judging whether the model meets the condition that the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum; if yes, go to step 14;
step 13: if the model does not meet the condition that the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, returning to the step 9, and optimizing the weight beta from L neurons of the hidden layer to neurons of the output layer by adopting a simulated annealing algorithm [ beta ]1,β2,…,βL]T；
Step 14: verifying the trained central air conditioner fault prediction and classification model by using the test sample;
step 15: judging whether the test sample meets the condition that the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum; if yes, go to step 16; otherwise, returning to the step 5;
step 16: and deploying a central air conditioner fault prediction and diagnosis classification model of a double intelligent algorithm on line.
FIG. 3 is a flow chart for solving the minimum fault classification rate of central air conditioner fault prediction and diagnosis by using a simulated annealing algorithm. In order to improve the identification precision and accuracy of the central air conditioner fault mode, the classification problem of central air conditioner fault prediction and diagnosis is converted into the optimization problem with the minimum model misclassification rate, and the global optimal solution with the minimum misclassification rate is solved through a simulated annealing algorithm. The designed central air conditioner fault prediction and diagnosis misclassification rate minimization problem solved based on the simulated annealing algorithm is as follows:
Figure BDA0002593168850000291
Figure BDA0002593168850000292
in which ξij(β*) Representing a central air conditioner fault prediction and diagnosis classification model classification function by a double intelligent algorithm; y isiRepresenting output layer neurons trained by neural networks and satisfying the minimum min y of output errori-ηi||；KFIs a penalty function.
Obtaining a central air conditioner fault prediction and diagnosis model parameter which minimizes the misclassification rate by solving a minimization problem, wherein the specific calculation process is as follows:
step 1: initializing relevant parameters of the simulated annealing algorithm: setting an initial value t of a temperature control parameter t0Setting Markov chain iteration times M under a temperature control parameter t, and selecting any initial solution beta in a feasible solution space based on an extreme learning machine central air-conditioning fault prediction and diagnosis model;
step 2: calculating the misclassification rate W of the central air-conditioning fault prediction and diagnosis model generated by betaerror(β)；
And step 3: generating beta by adopting random disturbance mode*Generating beta by random perturbation*：β*β + τ α, where α is a random vector whose constituent components are numbers between (0-1), τ being the step size of the iteration; calculating the sum of*Misclassification rate W of generated central air conditioner fault prediction and diagnosis modelerror(β*)；
And 4, step 4: judgment of Werror(β*)-Werror(β) is less than or equal to 0;
and 5: if Werror(β*)-WerrorIf beta is less than or equal to 0, then beta is used*Generating central air conditioner fault prediction and diagnosis modelEntering step 7;
step 6: if Werror(β*)-Werror(beta) > 0, then a bit at [0,1 ] is randomly generated]Random number in interval
Figure BDA0002593168850000301
Judgment of
Figure BDA0002593168850000302
Whether or not greater than
Figure BDA0002593168850000303
If it is
Figure BDA0002593168850000304
Then by beta*Generating a central air conditioner fault prediction and diagnosis classification model, and entering the step 7; if it is
Figure BDA0002593168850000305
The parameters of the central air conditioner fault prediction and diagnosis classification model are unchanged; the symbol exp () represents the exponent calculation;
and 7: judging whether the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum; if yes, entering step 8; otherwise, returning to the step 1, reducing the temperature control parameter t and continuing to operate;
and 8: deploying by beta online*And generating a central air conditioner fault prediction and diagnosis classification model with a double intelligent algorithm.
The central air conditioner fault detection and diagnosis method provided by the invention has the advantages of good learning stability in an off-line stage, high fault prediction and diagnosis precision in an on-line stage and the like, and meanwhile, the sensitivity of the central air conditioner fault detection and diagnosis method to abnormal data in data concentration is low. The design method of the invention has simple design, does not change the structure of the existing central air-conditioning system, has convenient operation and is convenient for construction and popularization.

Claims (9)

1. A double-intelligent-algorithm central air conditioner fault prediction and diagnosis method is characterized by comprising the following steps: the method comprises the following steps:
first, collecting siteCentral air-conditioning operation real-time data set Son-line；
Secondly, the real-time data set S in the step of one is collectedon-lineInputting the data into a pre-established central air conditioner fault prediction and diagnosis classification neural network model based on a simulated annealing algorithm and an extreme learning machine algorithm to judge the real-time data set S in the step oneon-lineWhether the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum or not is met;
real-time data set S in the third and second stepson-lineAnd if the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, outputting a fault prediction and diagnosis classification result of the central air conditioner.
2. The dual intelligent algorithm central air conditioner fault prediction and diagnosis method of claim 1, characterized in that: in the step two: if the real-time data set S in the step of "oneon-lineIf the fault prediction and diagnosis misclassification rate of the central air conditioner is not met to be minimum, the field central air conditioner operation data set S is usedon-lineAnd an initial data set S collected when a central air conditioner fault prediction and diagnosis classification model is establishedoff-lineAnd merging, and training a central air conditioner fault prediction and diagnosis classification neural network model.
3. The dual intelligent algorithm central air conditioner fault prediction and diagnosis method of claim 2, characterized in that: the method for constructing the central air-conditioning fault prediction and diagnosis classification neural network model comprises the following steps:
step 1: the historical operation data collected by the central air-conditioning heating/cooling heating and cooling cycle field sensor is used as an initial data set Soff-lineAnd the selected initial data set Soff-lineThe data comprises data with normal operation and data with failure operation;
step 2: judging whether the fault prediction and diagnosis classification neural network model meets the condition that the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum;
and step 3: if the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, entering a step 7;
and 4, step 4: if the fault prediction and diagnosis misclassification rate of the central air conditioner is not met, entering a step 6;
and 5: optimizing the central air-conditioning fault prediction and diagnosis classification model coefficient of the extreme learning machine by adopting a simulated annealing algorithm to obtain a new central air-conditioning fault prediction and diagnosis classification neural network model, and then judging whether the new central air-conditioning fault prediction and diagnosis classification neural network model meets the condition that the central air-conditioning fault prediction and diagnosis misclassification rate is minimum; if the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, entering
Step 6; otherwise, continuing optimization until the minimum misclassification rate is met;
and 7: and deploying a central air conditioner fault prediction and diagnosis classification model of a double intelligent algorithm on line.
4. The dual intelligent algorithm central air conditioning fault prediction and diagnosis method of claim 3, characterized in that: in step 2, the method for training the central air-conditioning fault prediction and diagnosis classification neural network model of the extreme learning machine specifically comprises the following steps:
step 1: the historical operation data collected by the central air-conditioning heating/cooling heating and cooling cycle field sensor is used as an initial data set Soff-lineAnd the selected sample comprises data of normal operation and data of fault operation;
step 2: constructing a single-layer feedforward neural network model consisting of 12 input neurons, 4 output neurons and L hidden layer neurons; wherein the input neurons are respectively at inhalation pressure x1Evaporator approach temperature x2The approach temperature x of the condenser3Return water temperature x of evaporator4Outlet water temperature x of evaporator5Condenser return water temperature x6Outlet water temperature x of condenser7Water supply pressure x of water separator8Water collector return pressure x9Current x10Voltage x11Heating/cooling and refrigerating capacity x of central air conditioner12(ii) a The output neurons are respectively of the normal operation class of the central air conditionerType y1Evaporator failure type y2Type of condenser failure y3And compressor failure type y4；
And step 3: normalizing the 12 input neurons;
and 4, step 4: initial data set Soff-lineDivided into training sample sets StrainAnd a test sample set Stest(ii) a Wherein the training sample set StrainThe data comprises data of a normal operation type and data of an operation fault type, and the total number of the N discrete training samples
Figure FDA0002593168840000021
Wherein xi＝[xi1,xi2,…,xi12]T∈R12(ii) a Component xijA j-th input neuron representing an i-th sample, and i ═ 1,2, …, N; j ═ 1,2, …, 12; x is the number ofi11 st input neuron inspiratory pressure, x, representing the ith samplei2The 2 nd input neuron evaporator approach temperature, x, representing the ith samplei3The 3 rd input neuron condenser approach temperature, x, representing the ith samplei44 th input neuron evaporator Return Water temperature, x, representing the ith samplei5The evaporator leaving water temperature, x, of the 5 th input neuron representing the ith samplei6Condenser return water temperature, x, of the 6 th input neuron representing the ith samplei7Condenser outlet water temperature, x, of the 7 th input neuron representing the ith samplei8Water supply pressure, x, for the 8 th input neuron trap representing the ith samplei9The 9 th input neuron collector return water pressure, x, representing the ith samplei1010 th input neuron current, x, representing the ith samplei11The 11 th input neuron voltage, x, representing the ith samplei12The 12 th input neuron central air-conditioning heating/cooling heating and cooling capacity representing the ith sample; x is the number ofiAs an input to a single-layer feedforward neural network model; etai＝[ηi1,ηi2,ηi3,ηi4]T∈R4Indicates the results of normal operation contained in the sample andfault operation result data; component ηi1The ith sample is the normal operation type of the central air conditioner, etai2Indicating the i-th sample as evaporator fault type, etai3Indicating the ith sample as the type of condenser fault, ηi4Indicating the ith sample as a compressor fault type; symbol [ 2 ]]TRepresents a transpose of a matrix;
and 5: selecting L as the number of hidden layer neurons;
step 6: adopting a random method to generate the weight value w of the kth neuron of the hidden layer and the 12 neurons of the input layerk＝[wk1,wk2,…,wk12]T(ii) a Generating a bias value b of the kth neuron of the hidden layer by adopting a random methodk，k＝1,2,…,L；
And 7: selecting an excitation function g (w) of a central air-conditioning fault prediction and classification neural network model of the extreme learning machinek·xi+bk) And i is 1,2, …, N, k is 1,2, …, L;
and 8: using the weight value w calculated in step 61,…,wLAnd an offset value b1,...,bLAnd the excitation function in the step 7, and calculating the hidden layer output matrix H ═ H (w)1,...,wL,b1,...,bL,x1,…,xN)；
And step 9: calculating the weight beta from L neurons of the hidden layer to neurons of the output layer by using the hidden layer output matrix H obtained in the step 81,β2,…,βL]T(ii) a Wherein beta isk＝[βk1,βk2,βk3,βk4]TThe weight k representing the k-th neuron of the hidden layer to 4 neurons of the output layer is 1,2, …, L;
step 10: judging whether the beta in the step 9 meets the requirement of minimum output error of the single-layer feedforward neural network, if so, entering a step 11; otherwise, returning to the step 5;
step 11: outputting a central air conditioner fault prediction and classification model based on an extreme learning machine;
step 12: judging whether the central air-conditioning fault prediction and classification model based on the extreme learning machine in the step 11 meets the condition that the central air-conditioning fault prediction and diagnosis misclassification rate is minimum; if yes, entering step 14, otherwise, directly entering step 13;
step 13: if the model does not meet the condition that the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum, returning to the step 9, and optimizing the weight beta from L neurons of the hidden layer to neurons of the output layer by adopting a simulated annealing algorithm [ beta ]1,β2,…,βL]T；
Step 14: verifying the trained central air conditioner fault prediction and classification model by using the test sample;
step 15: judging whether the test sample meets the condition that the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum; if yes, go to step 16; otherwise, returning to the step 5.
5. The dual intelligent algorithm central air conditioner fault prediction and diagnosis method of claim 4, characterized in that: step 13, optimizing the weights beta from L neurons in the hidden layer to neurons in the output layer by adopting a simulated annealing algorithm1,β2,…,βL]TThe method comprises the following specific steps:
step 1: initializing relevant parameters of a simulated annealing algorithm, and selecting any initial solution beta in a feasible solution space based on a central air conditioner fault prediction and diagnosis model of an extreme learning machine;
step 2: calculating the misclassification rate W of the central air-conditioning fault prediction and diagnosis model generated by betaerror(β)；
And step 3: randomly generating beta*Calculating by beta*Misclassification rate W of generated central air conditioner fault prediction and diagnosis modelerror(β*)；
And 4, step 4: judgment of Werror(β*)-Werror(β) is less than or equal to 0;
and 5: if Werror(β*)-WerrorIf beta is less than or equal to 0, then beta is used*Generating a central air conditioner fault prediction and diagnosis model, and entering the step 7;
step 6: if Werror(β*)-Werror(beta) > 0, then a bit at [0,1 ] is randomly generated]Random number in interval
Figure FDA0002593168840000031
Judgment of
Figure FDA0002593168840000032
Whether or not greater than
Figure 1
If it is
Figure FDA0002593168840000034
Then by beta*Generating a central air conditioner fault prediction and diagnosis classification model, and entering the step 7; if it is
Figure FDA0002593168840000035
The parameters of the central air conditioner fault prediction and diagnosis classification model are unchanged; the symbol exp () represents the exponent calculation;
and 7: judging whether the fault prediction and diagnosis misclassification rate of the central air conditioner is minimum; if yes, entering step 8; otherwise, returning to the step 1, and reducing the temperature control parameter t to continue the operation.
6. The dual intelligent algorithm central air conditioner fault prediction and diagnosis method of claim 5, characterized in that: in step 2, the misclassification rate W of the central air conditioner fault prediction and diagnosis model generated by beta is calculatederrorThe method (β) is specifically as follows:
the quality evaluation of the central air-conditioning fault prediction and diagnosis classification model is evaluated by adopting fault misclassification rate conversion, and a misclassification rate calculation method is described by taking a training sample set as an example;
training sample set StrainThe data comprises normal operation data and operation fault data, and the total number of the N discrete training samples
Figure FDA0002593168840000041
W1Is to beThe samples output by the central air-conditioning fault prediction and diagnosis model as the normal operation types are classified into fault types by mistake, or the samples output by the central air-conditioning fault prediction and diagnosis model as the fault types are classified into a set of normal operation types by mistake;
W2the method comprises the steps of classifying a sample which is output as a fault state by a model into a fault type, but not belonging to a data set of a correct fault type sample; if the original output type is evaporator fault but misclassification is condenser fault or compressor fault, the original output type is condenser fault but misclassification is evaporator fault or compressor fault, and the original output type is compressor fault but misclassification is evaporator fault or condenser fault;
W3is a set of classification rates for which the output type is correct, and satisfies
Figure FDA0002593168840000042
Wherein the symbols
Figure FDA0002593168840000043
Is a direct sum operation symbol;
the minimum misclassification transformation of the central air-conditioning fault prediction and diagnosis is as follows:
Figure FDA0002593168840000044
Figure FDA0002593168840000045
in which ξij(beta) represents a central air conditioner fault prediction and diagnosis classification model classification function by a dual intelligent algorithm; y isiRepresenting output layer neurons trained by neural networks, yi＝[yi1,yi2,yi3,yi4]TI is 1,2, …, N, and satisfies the output error min yi-ηi||。
7. The dual intelligent algorithm central air conditioner fault prediction and diagnosis method of claim 4, characterized in that: in step 3, the 12 input neuron parameters are normalized to 0-1 numerical values, and the specific normalization method is as follows:
(1) inspiratory pressure parameter x1Is normalized to obtain
Figure FDA0002593168840000051
Figure FDA0002593168840000052
Wherein
Figure FDA0002593168840000053
The air suction pressure under the working condition is designed for the central air-conditioning system,
Figure 2
is a normalized inspiratory pressure parameter;
(2) evaporator approach temperature parameter x2Is normalized to obtain
Figure FDA0002593168840000055
Figure FDA0002593168840000056
Wherein
Figure FDA0002593168840000057
The evaporator approaches the temperature under the design working condition of the central air-conditioning system,
Figure FDA0002593168840000058
the evaporator approach temperature parameter after normalization;
(3) condenser approach temperature x3Is normalized to obtain
Figure FDA0002593168840000059
Figure FDA00025931688400000510
Wherein
Figure FDA00025931688400000511
The approach temperature of the condenser under the working condition is designed for the central air-conditioning system;
Figure FDA00025931688400000512
the normalized condenser approach temperature parameter is obtained;
(4) return water temperature x of evaporator4Is normalized to obtain
Figure FDA00025931688400000513
Figure FDA00025931688400000514
Wherein
Figure FDA00025931688400000515
The return water temperature of the evaporator under the working condition is designed for the central air-conditioning system;
Figure FDA00025931688400000516
the normalized return water temperature parameter of the evaporator is obtained;
(5) water outlet temperature x of evaporator5Is normalized to obtain
Figure FDA00025931688400000517
Figure FDA00025931688400000518
Wherein
Figure FDA00025931688400000519
The outlet water temperature of the evaporator under the working condition is designed for the central air-conditioning system;
Figure FDA00025931688400000520
the normalized outlet water temperature parameter of the evaporator is obtained;
(6) condenser return water temperature x6Is normalized to obtain
Figure FDA00025931688400000521
Figure FDA00025931688400000522
Wherein
Figure FDA00025931688400000523
The return water temperature of the condenser under the working condition is designed for the central air-conditioning system;
Figure FDA00025931688400000524
the normalized condenser return water temperature parameter is obtained;
(7) condenser outlet water temperature x7Is normalized to obtain
Figure FDA00025931688400000525
Figure FDA00025931688400000526
Wherein
Figure FDA00025931688400000527
Designing the outlet water temperature of the condenser under the working condition for the central air-conditioning system;
Figure FDA00025931688400000528
the water outlet temperature parameter is the normalized condenser water outlet temperature parameter;
(8) water supply pressure x of water separator8Is normalized to obtain
Figure FDA00025931688400000529
Figure FDA00025931688400000530
Wherein
Figure FDA00025931688400000531
The water supply pressure of the water distributor under the working condition is designed for the central air-conditioning system;
Figure 3
the water supply pressure parameter of the water separator after normalization;
(9) water collector return pressure x9Is normalized to obtain
Figure FDA00025931688400000533
Figure FDA00025931688400000534
Wherein
Figure FDA00025931688400000535
The backwater pressure of the water collector under the working condition is designed for the central air-conditioning system;
Figure FDA00025931688400000536
the water pressure parameter is the normalized water collector return water pressure parameter;
(10) current x10Is normalized to obtain
Figure FDA0002593168840000061
Figure FDA0002593168840000062
Wherein
Figure FDA0002593168840000063
Rated current for safe operation of the central air-conditioning system;
Figure FDA0002593168840000064
is a normalized current parameter;
(11) voltage x11Is normalized to obtain
Figure FDA0002593168840000065
Figure FDA0002593168840000066
Wherein
Figure FDA0002593168840000067
Rated voltage for safe operation of the central air-conditioning system;
Figure FDA0002593168840000068
the normalized voltage parameter is obtained;
(12) heating and cooling capacity x of central air conditioner12Is normalized to obtain
Figure FDA0002593168840000069
Figure FDA00025931688400000610
Wherein
Figure FDA00025931688400000611
The calculation method is as follows for the nominal heating capacity of the central air conditioner:
Figure FDA00025931688400000612
wherein K1Correcting coefficient of heating capacity or cooling capacity of the central air conditioning unit; k2A defrosting correction coefficient of the central air conditioning unit; k, the comprehensive heat transfer coefficient of the building; t isnHeating and cooling indoor design temperature, TpCalculating the outdoor temperature in the heating season or the refrigerating season;
Figure FDA00025931688400000613
heating and refrigerating capacity parameters of the normalized central air conditioner;
where the symbol | | | represents the modulus of the vector.
8. The dual intelligent algorithm central air conditioner fault prediction and diagnosis method of claim 4, characterized in that: step 8 calculates the hidden layer output matrix H ═ H (w)1,...,wL,b1,...,bL,x1,...,xN)：
Figure FDA00025931688400000614
Wherein x1＝[x11,x12,…,x112]TRepresents the 1 st sample parameter in the training dataset: the air-conditioning system comprises an air suction pressure, an evaporator approach temperature, a condenser approach temperature, an evaporator return water temperature, an evaporator outlet water temperature, a condenser return water temperature, a condenser outlet water temperature, a water distributor supply pressure, a water collector return water pressure, current, voltage and central air-conditioning heating and refrigerating capacity; the 12 parameters are used as input layer parameters of a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
wherein xN＝[xN1,xN2,…,xN12]TRepresents the nth sample parameter in the training dataset: the air-conditioning system comprises an air suction pressure, an evaporator approach temperature, a condenser approach temperature, an evaporator return water temperature, an evaporator outlet water temperature, a condenser return water temperature, a condenser outlet water temperature, a water distributor supply pressure, a water collector return water pressure, current, voltage and central air-conditioning heating and refrigerating capacity; the 12 parameters are used as input layer parameters of a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
wherein w1＝[w11,w12,…,w112]Representing the weight values of the 1 st neuron of the hidden layer and the 12 neurons of the input layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm, and the component w11Representing the weight value of the 1 st neuron of the hidden layer and the inspiration pressure neuron of the input layer; w is a12Representing the weight value of the 1 st neuron of the hidden layer and the approximate temperature neuron of the evaporator of the input layer; w is a13Representing the weight value of the 1 st neuron of the hidden layer and the input layer condenser approaching temperature neuron; w is a14Representing the weight values of the 1 st neuron of the hidden layer and the return water temperature neuron of the evaporator of the input layer; w is a15Representing the weight values of the 1 st neuron of the hidden layer and the water outlet temperature neuron of the input layer evaporator; w is a16Representing the weight values of the 1 st neuron of the hidden layer and the return water temperature neuron of the condenser of the input layer; w is a17Representing the weight values of the 1 st neuron of the hidden layer and the water outlet temperature neuron of the input layer condenser; w is a18Representing the weight values of the 1 st neuron of the hidden layer and the water supply pressure neuron of the input layer water separator; w is a19Representing the weight value of the 1 st neuron of the hidden layer and the backwater pressure neuron of the water collector of the input layer; w is a110Representing the weight value of the 1 st neuron of the hidden layer and the current neuron of the input layer; w is a111Representing the weight values of the 1 st neuron of the hidden layer and the input layer voltage neuron; w is a112Representing the weight values of the 1 st neuron of the hidden layer and the heating and cooling capacity neuron of the central air conditioner of the input layer;
wherein wL＝[wL1,wL2,…,wL12]Representing the weight values of the Lth neuron of the hidden layer and the 12 neurons of the input layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm, and the component wL1Representing the weight value of the Lth neuron of the hidden layer and the inspiration pressure neuron of the input layer; w is aL2The weight value of the Lth neuron of the hidden layer and the approximate temperature neuron of the evaporator of the input layer is represented; w is aL3Representing the weight value of the Lth neuron of the hidden layer and the input layer condenser approaching temperature neuron; w is aL4Representing the weight values of the Lth neuron of the hidden layer and the return water temperature neuron of the evaporator of the input layer; w is aL5Representing the weight values of the L-th neuron of the hidden layer and the water outlet temperature neuron of the input layer evaporator; w is aL6Representing the weight values of the Lth neuron of the hidden layer and the return water temperature neuron of the condenser of the input layer; w is aL7Representing the weight values of the Lth neuron of the hidden layer and the water outlet temperature neuron of the input layer condenser; w is aL8Representing the weight value of the L-th neuron of the hidden layer and the water supply pressure neuron of the input layer water separator; w is aL9Representing the weight value of the Lth neuron of the hidden layer and the backwater pressure neuron of the water collector of the input layer; w is aL10Representing hidden layer 1 st neuron and input layer current neuronA weight value; w is aL11Representing the weight values of the Lth neuron of the hidden layer and the voltage neuron of the input layer; w is aL12Representing the weight values of the Lth neuron of the hidden layer and the heating and cooling capacity neuron of the central air conditioner of the input layer; "·" denotes an inner product operation; with wL·x1The description is given for the sake of example: w is aL·x1＝wL1×x11+wL2×x12+…wL12×x112∈R；
Excitation function g (w)k·xi+bk) Mapping a linear space containing the air suction pressure, the approach temperature of an evaporator, the approach temperature of a condenser, the return water temperature of the evaporator, the outlet water temperature of the evaporator, the return water temperature of the condenser, the outlet water temperature of the condenser, the water supply pressure of a water distributor, the return water pressure of a water collector, current, voltage and the heating and refrigerating capacity of the central air conditioner in a central air conditioner fault prediction and diagnosis neural network model based on an extreme learning machine algorithm into a linear space containing L hidden layer characteristics in the extreme learning machine; the specific selection method of the excitation function is as follows:
trigonometric function: g (w)k·xi+bk)＝cos(wk·xi+bk)；
Gaussian function:
Figure FDA0002593168840000081
sigmoid function:
Figure FDA0002593168840000082
hyperbolic sine function:
Figure FDA0002593168840000083
radial basis function: g (w)k·xi+bk)＝exp(-bk·||xi-wk||)。
9. Dual intelligent algorithmic center space as in claim 4The fault adjustment prediction and diagnosis method is characterized by comprising the following steps: and step 9: according to β ═ H+Y calculates the weight from L neurons in the hidden layer to the output layer neuron, and the hidden layer output weight beta is [ beta ]1,β2,…,βL]；
Wherein beta is1＝[β11,β12,β13,β14]TMiddle component beta11The central air conditioner fault prediction and diagnosis neural network model based on the extreme learning machine algorithm is characterized by comprising 1 st neuron and 1 st output layer neuron of a hidden layer, wherein the 1 st neuron and the 1 st output layer neuron are respectively connected with a central air conditioner; component beta12Representing the weight values of the evaporator fault types of the 1 st neuron and the 2 nd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta13Representing the weight values of the condenser fault types of the 1 st neuron and the 3 rd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta14Representing the weight values of compressor fault types of 1 st neuron and 4 th output layer neuron of a hidden layer in a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
wherein beta is2＝[β21,β22,β23,β24]TMiddle component beta21The central air conditioner fault prediction and diagnosis neural network model based on the extreme learning machine algorithm is characterized by comprising the following steps of (1) representing the weight value of the normal operation type of the central air conditioner of the hidden layer 2 th neuron and the hidden layer 1 st output layer neuron; component beta22Representing the weight values of the evaporator fault types of the neurons of the No. 2 and the No. 2 output layers of the hidden layers in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta23Representing the weight values of the condenser fault types of the 2 nd neuron and the 3 rd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta24Representing 2 nd nerve of hidden layer in central air-conditioning fault prediction and diagnosis neural network model based on extreme learning machine algorithmThe weighted value of the compressor fault type of the neuron and the 4 th output layer neuron;
wherein beta isL＝[βL1,βL2,βL3,βL4]TMiddle component beta11The central air conditioner fault prediction and diagnosis neural network model based on the extreme learning machine algorithm is characterized by comprising the following steps of (1) representing the weight value of the normal operation type of the central air conditioner of the hidden layer Lth neuron and the 1 st output layer neuron; component beta12Representing the weight values of the evaporator fault types of the L-th neuron and the 2 nd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta13Representing the weight values of the condenser fault types of the Lth neuron and the 3 rd output layer neuron of the hidden layer in the central air-conditioning fault prediction and diagnosis neural network model based on the extreme learning machine algorithm; component beta14Representing the weight values of compressor fault types of the Lth neuron and the 4 th output layer neuron of a hidden layer in a central air-conditioning fault prediction and diagnosis neural network model based on an extreme learning machine algorithm;
H+output matrix H ═ H (w) for hidden layers1,...,wL,b1,...,bL,x1,...,xN) Moore-Penrose generalized inverse matrix of (1); when H is presentTH is not singular: h+＝(HTH)-1HTOr H+＝HT(HTH)-1(ii) a Wherein ()-1Represents the inverse of the matrix ()TRepresents a transpose of a matrix;
Y＝[y1,y2,…,yN]Tzero-error approximation of N discrete training samples for training
Figure FDA0002593168840000091
Of the neural network, wherein yi＝[yi1,yi2,yi3,yi4]T，i＝1,2,…,N；
At y1Middle, component y11Representing extreme learning machine neural network model with zero error approximation from the 1 st sampleThe 1 st neuron output after training is of the normal operation type of the central air conditioner; component y12Representing that the 2 nd neuron output after the 1 st training sample is subjected to extreme learning machine neural network model training is an evaporator fault type by zero error approximation; component y13Representing that the 3 rd neuron output after the training of the neural network model of the extreme learning machine is carried out on the 1 st training sample by using zero error approximation is a condenser fault type; component y14Representing that the 4 th neuron output after the training of the neural network model of the extreme learning machine is carried out on the 1 st training sample by using zero error approximation is a compressor fault type;
at y2Middle, component y21Representing that the 1 st neuron output after the 2 nd sample is subjected to extreme learning machine neural network model training is a normal operation type of the central air conditioner by using zero error approximation; component y22Representing that the 2 nd neuron output after the training of the neural network model of the extreme learning machine is carried out on the 2 nd training sample by zero error approximation is an evaporator fault type; component y23Representing that the 3 rd neuron output after the 2 nd training sample is subjected to extreme learning machine neural network model training is a condenser fault type by zero error approximation; component y24Representing that the 4 th neuron output after the 2 nd training sample is subjected to extreme learning machine neural network model training is a compressor fault type by zero error approximation;
at yNMiddle, component yN1Representing that the 1 st neuron output after the training of the neural network model of the extreme learning machine is carried out on the Nth training sample by using zero error approximation is the normal operation type of the central air conditioner; component yN2Representing that the Nth neuron output after the Nth training sample is subjected to extreme learning machine neural network model training by using zero error approximation is an evaporator fault type; component yN3Representing that the 3 rd neuron output after the training of the neural network model of the extreme learning machine is carried out on the Nth training sample by using zero error approximation is a condenser fault type; component yN4Representing the 4 th neuron output after the training of the neural network model of the extreme learning machine is carried out on the Nth training sample by zero error approximationA compressor failure type;
yithe calculation method is
Figure FDA0002593168840000092
And the minimum min y of the error output by the neural network model is satisfiedi-ηi||，i＝1,2,…,N，k＝1,2,…,L。
CN202010702092.3A 2020-07-21 2020-07-21 Central air conditioner fault prediction and diagnosis method adopting double intelligent algorithms Active CN111981635B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010702092.3A CN111981635B (en) 2020-07-21 2020-07-21 Central air conditioner fault prediction and diagnosis method adopting double intelligent algorithms

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010702092.3A CN111981635B (en) 2020-07-21 2020-07-21 Central air conditioner fault prediction and diagnosis method adopting double intelligent algorithms

Publications (2)

Publication Number Publication Date
CN111981635A true CN111981635A (en) 2020-11-24
CN111981635B CN111981635B (en) 2021-12-31

Family

ID=73439321

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010702092.3A Active CN111981635B (en) 2020-07-21 2020-07-21 Central air conditioner fault prediction and diagnosis method adopting double intelligent algorithms

Country Status (1)

Country Link
CN (1) CN111981635B (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112556088A (en) * 2020-11-26 2021-03-26 珠海格力电器股份有限公司 Air conditioner and detection method thereof
CN112766327A (en) * 2021-01-05 2021-05-07 格力电器（武汉）有限公司 Air conditioner fault prediction method, electronic equipment and storage medium
CN112966399A (en) * 2021-04-15 2021-06-15 苏州大学张家港工业技术研究院 Pulse tube refrigerator working condition prediction method and system based on machine learning
CN113158546A (en) * 2021-03-11 2021-07-23 南京信息工程大学 Air conditioning unit equipment fault prediction method based on HSMM
CN116678072A (en) * 2023-06-05 2023-09-01 耀昶嵘相变材料科技(广东)有限公司 Fault processing method and terminal of central air conditioning system and central air conditioning system

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018092258A1 (en) * 2016-11-18 2018-05-24 三菱電機株式会社 Air conditioner and air-conditioning system
CN108507117A (en) * 2017-10-13 2018-09-07 上海智容睿盛智能科技有限公司 A kind of Air-conditioning system sensor method for diagnosing faults based on wavelet neural network
CN109948778A (en) * 2019-02-28 2019-06-28 上海理工大学 A kind of refrigeration equipment trouble-shooter and system
CN110160210A (en) * 2019-05-15 2019-08-23 北京上格云技术有限公司 Fault detection method, device, storage medium and the electronic equipment of fan coil
CN110677290A (en) * 2019-09-26 2020-01-10 国网安徽省电力有限公司信息通信分公司 Electric power communication operation and maintenance risk early warning method based on annealing algorithm and neural network model
CN111174370A (en) * 2018-11-09 2020-05-19 珠海格力电器股份有限公司 Fault detection method and device, storage medium and electronic device

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018092258A1 (en) * 2016-11-18 2018-05-24 三菱電機株式会社 Air conditioner and air-conditioning system
CN108507117A (en) * 2017-10-13 2018-09-07 上海智容睿盛智能科技有限公司 A kind of Air-conditioning system sensor method for diagnosing faults based on wavelet neural network
CN111174370A (en) * 2018-11-09 2020-05-19 珠海格力电器股份有限公司 Fault detection method and device, storage medium and electronic device
CN109948778A (en) * 2019-02-28 2019-06-28 上海理工大学 A kind of refrigeration equipment trouble-shooter and system
CN110160210A (en) * 2019-05-15 2019-08-23 北京上格云技术有限公司 Fault detection method, device, storage medium and the electronic equipment of fan coil
CN110677290A (en) * 2019-09-26 2020-01-10 国网安徽省电力有限公司信息通信分公司 Electric power communication operation and maintenance risk early warning method based on annealing algorithm and neural network model

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112556088A (en) * 2020-11-26 2021-03-26 珠海格力电器股份有限公司 Air conditioner and detection method thereof
CN112556088B (en) * 2020-11-26 2022-08-02 珠海格力电器股份有限公司 Air conditioner and detection method thereof
CN112766327A (en) * 2021-01-05 2021-05-07 格力电器（武汉）有限公司 Air conditioner fault prediction method, electronic equipment and storage medium
CN112766327B (en) * 2021-01-05 2024-05-24 格力电器（武汉）有限公司 Air conditioner fault prediction method, electronic equipment and storage medium
CN113158546A (en) * 2021-03-11 2021-07-23 南京信息工程大学 Air conditioning unit equipment fault prediction method based on HSMM
CN113158546B (en) * 2021-03-11 2023-05-09 南京信息工程大学 Fault prediction method of air conditioning unit equipment based on HSMM
CN112966399A (en) * 2021-04-15 2021-06-15 苏州大学张家港工业技术研究院 Pulse tube refrigerator working condition prediction method and system based on machine learning
CN112966399B (en) * 2021-04-15 2023-08-22 苏州大学张家港工业技术研究院 Pulse tube refrigerator working condition prediction method and system based on machine learning
CN116678072A (en) * 2023-06-05 2023-09-01 耀昶嵘相变材料科技(广东)有限公司 Fault processing method and terminal of central air conditioning system and central air conditioning system
CN116678072B (en) * 2023-06-05 2024-05-10 耀昶嵘相变材料科技(广东)有限公司 Fault processing method and terminal of central air conditioning system and central air conditioning system

Also Published As

Publication number Publication date
CN111981635B (en) 2021-12-31

Similar Documents

Publication Publication Date Title
CN111981635B (en) Central air conditioner fault prediction and diagnosis method adopting double intelligent algorithms
CN111832814B (en) Air pollutant concentration prediction method based on graph attention mechanism
CN108734360B (en) Correction-based ELM prediction model multidimensional telemetering data intelligent interpretation method
CN107817404B (en) Portable metering automation terminal fault diagnosis device and diagnosis method thereof
CN110555230B (en) Rotary machine residual life prediction method based on integrated GMDH framework
CN106156401B (en) Multi-combination classifier based data driving system state model online identification method
CN109213127A (en) A kind of HVAC system gradual failure diagnostic method based on deep learning
CN111461413B (en) Highway road surface performance detecting system
CN111140986A (en) Operating state detection method and device of air conditioning system, storage medium and air conditioner
CN111191726B (en) Fault classification method based on weak supervision learning multilayer perceptron
CN110363334B (en) Grid line loss prediction method of photovoltaic grid connection based on gray neural network model
CN115358347B (en) Method for predicting remaining life of intelligent electric meter under different subsystems
CN111639823A (en) Building cold and heat load prediction method constructed based on feature set
CN109948920B (en) Electric power market settlement data risk processing method based on evidence theory
CN110503153A (en) Photovoltaic system method for diagnosing faults based on differential evolution algorithm and support vector machines
CN115455746B (en) Nuclear power device operation monitoring data anomaly detection and correction integrated method
CN110072205A (en) A kind of layering aggregation method for wireless sense network anomaly data detection
CN115222104A (en) Intelligent substation secondary equipment state evaluation method based on extreme learning machine
CN113447879A (en) Electric energy meter metering error prediction method, system and device based on multi-stress influence
CN114912339A (en) Multi-factor temperature prediction method for primary equipment of transformer substation
CN114819102A (en) GRU-based air conditioning equipment fault diagnosis method
CN113486950B (en) Intelligent pipe network water leakage detection method and system
TW202119298A (en) Solar panel power system abnormal diagnosis and analysis device and method based on fhmm (factorial hidden markov model) and power generation prediction
CN112819087B (en) Method for detecting abnormality of BOD sensor of outlet water based on modularized neural network
Correa-Jullian et al. Assessment of deep learning algorithms for fault diagnosis in solar thermal systems

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant