CN113654143A - Air-conditioning water system pump valve linkage control method based on neural network - Google Patents

Air-conditioning water system pump valve linkage control method based on neural network Download PDF

Info

Publication number
CN113654143A
CN113654143A CN202110946317.4A CN202110946317A CN113654143A CN 113654143 A CN113654143 A CN 113654143A CN 202110946317 A CN202110946317 A CN 202110946317A CN 113654143 A CN113654143 A CN 113654143A
Authority
CN
China
Prior art keywords
neural network
air
network
branch
layer
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
CN202110946317.4A
Other languages
Chinese (zh)
Other versions
CN113654143B (en
Inventor
高波
倪吉
袁中原
于佳佳
朱晓玥
周耀鹏
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.)
Sichuan Institute of Building Research
Original Assignee
Sichuan Institute of Building Research
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 Sichuan Institute of Building Research filed Critical Sichuan Institute of Building Research
Priority to CN202110946317.4A priority Critical patent/CN113654143B/en
Publication of CN113654143A publication Critical patent/CN113654143A/en
Application granted granted Critical
Publication of CN113654143B publication Critical patent/CN113654143B/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
    • F24F5/00—Air-conditioning systems or apparatus not covered by F24F1/00 or F24F3/00, e.g. using solar heat or combined with household units such as an oven or water heater
    • F24F5/0003—Exclusively-fluid systems
    • 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
    • F24F11/70—Control systems characterised by their outputs; Constructional details thereof
    • F24F11/80—Control systems characterised by their outputs; Constructional details thereof for controlling the temperature of the supplied air
    • F24F11/83—Control systems characterised by their outputs; Constructional details thereof for controlling the temperature of the supplied air by controlling the supply of heat-exchange fluids to heat-exchangers
    • F24F11/84—Control systems characterised by their outputs; Constructional details thereof for controlling the temperature of the supplied air by controlling the supply of heat-exchange fluids to heat-exchangers using valves
    • 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
    • F24F11/70—Control systems characterised by their outputs; Constructional details thereof
    • F24F11/80—Control systems characterised by their outputs; Constructional details thereof for controlling the temperature of the supplied air
    • F24F11/83—Control systems characterised by their outputs; Constructional details thereof for controlling the temperature of the supplied air by controlling the supply of heat-exchange fluids to heat-exchangers
    • F24F11/85—Control systems characterised by their outputs; Constructional details thereof for controlling the temperature of the supplied air by controlling the supply of heat-exchange fluids to heat-exchangers using variable-flow pumps
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02B—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO BUILDINGS, e.g. HOUSING, HOUSE APPLIANCES OR RELATED END-USER APPLICATIONS
    • Y02B30/00—Energy efficient heating, ventilation or air conditioning [HVAC]
    • Y02B30/70—Efficient control or regulation technologies, e.g. for control of refrigerant flow, motor or heating

Abstract

The invention discloses an air-conditioning water system pump valve linkage control method based on a neural network, which comprises the following steps: establishing and training a GA-BP network through actually measured data of a pump valve of an air-conditioning water system; calculating target flow of each branch required by air conditioner water system balance according to real-time load of the air conditioner; solving through a GA-BP network according to the target flow of each branch to obtain the opening of a balance valve of each branch and the current frequency of a water pump; and adjusting the balance valve of each branch of the air-conditioning water system according to the opening of the balance valve of each branch, and controlling the water pump according to the current frequency of the water pump to complete the linkage control of the pump valve. The invention uses the GA-BP network combining the GA genetic model and the BP neural network as an operation core, rapidly solves the opening degree of each branch balancing valve, solves the problems of slow response and inaccurate adjustment in the prior art, does not need to repeatedly adjust and teach to generate unnecessary power consumption, and can greatly reduce the whole power consumption by setting the proper current frequency of the water pump.

Description

Air-conditioning water system pump valve linkage control method based on neural network
Technical Field
The invention relates to the technical field of central air conditioners, in particular to an air conditioner water system pump valve linkage control method based on a neural network.
Background
During the operation and use of the central air conditioner, the load of the central air conditioner changes with the change of factors such as outdoor environment, indoor personnel, heat sources and the like. Under the premise that the temperature difference between the supplied water and the returned water is not changed, when the load of the air conditioner is changed, if the water flow of each user branch of the air conditioner water system is not adjusted in time, the phenomenon of uneven cooling and heating at each user can be caused, and dynamic hydraulic imbalance is caused.
For the operation regulation of an air-conditioning water system, the common method is to control the speed by frequency conversion of a water pump and regulate the opening of a branch valve of each user. The variable frequency speed regulation of the water pump can control the overall water flow of an air-conditioning water system; and the operation flow of each user branch can be accurately adjusted by adjusting the opening degree of the valve at each user branch.
The dynamic hydraulic imbalance phenomenon of the air-conditioning water system can be improved to a certain extent only by singly using any one of the two adjusting modes; the existing pump and valve linkage technology generally takes programmable logic devices such as a PLC (programmable logic controller) and the like as a core, establishes a control algorithm by using a traditional control equation, and interlocks and interconnects a water pump and a valve through an electric control loop, and is also limited by a plurality of branches and a complex pipe network of a central air-conditioning water system, so that the problem of difficulty in equation solution occurs, the system has large power consumption, slow response and inaccurate adjustment.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides the air-conditioning water system pump valve linkage control method based on the neural network, and solves the problems of large power consumption, slow response and inaccurate adjustment of the existing pump valve linkage technology system.
In order to achieve the purpose of the invention, the invention adopts the technical scheme that:
a pump-valve linkage control method of an air-conditioning water system based on a neural network comprises the following steps:
s1, establishing and training a GA-BP network through actually measured data of a pump valve of an air-conditioning water system;
s2, calculating target flow of each branch required by air conditioner water system balance according to real-time load of the air conditioner;
s3, solving through a GA-BP network according to the target flow of each branch to obtain the opening of each branch balance valve and the current frequency of the water pump;
and S4, adjusting the balance valve of each branch of the air-conditioning water system according to the opening of the balance valve of each branch, and controlling the water pump according to the current frequency of the water pump to complete the linkage control of the pump and the valve.
The invention has the beneficial effects that: the GA-BP network combining the GA genetic model and the BP neural network is used as an operation core, the opening degree of each branch balance valve can be rapidly solved under the condition of the target flow rate of each branch of the known air-conditioning water system and the current frequency of the water pump, and the problems of slow response and inaccurate adjustment of the pump valve linkage in the prior art are solved; the unidirectional water pump current frequency traversal optimization searching method also has the advantages of rapidness and effectiveness; benefit from the quick accurate operation of pump valve linkage parameter, the pump valve linkage need not retrace repeatedly and educates and produce unnecessary consumption, through the settlement of appropriate water pump current frequency, compares prior art but the whole consumption of greatly reduced.
Further, the step S1 includes the following sub-steps:
s11, establishing a three-layer BP neural network;
s12, setting boundary conditions of a three-layer BP neural network and a GA genetic model;
s13, randomly generating opening values of all branch balance valves and water pump current frequencies of M groups of air-conditioning water systems through a computer, controlling the air-conditioning water systems by using the opening values and the water pump current frequencies, actually measuring to obtain corresponding flow values of all branches of the M air-conditioning water systems, and taking the opening values of all branch balance valves, the water pump current frequencies and the flow values of all branches of the air-conditioning water systems of the M groups as sample sets;
s14, randomly dividing the sample set into a training set, a testing set and a verification set;
s15, according to the training set, calculating the optimal weight and threshold of the three-layer BP neural network through a GA genetic model, and completing the establishment of the GA-BP network;
and S16, performing iterative optimization and test verification on the GA-BP network according to the training set, the test set and the verification set to finish the training of the GA-BP network.
The beneficial effects of the above further scheme are: the BP neural network is a forward network based on error back propagation, has very strong nonlinear mapping capability, is suitable for solving the opening degree of each branch balance valve of an air-conditioning water system, has the efficiency far higher than that of a traditional air-conditioning water system control method based on a control equation, but the accuracy of the BP neural network depends on a training mode, and the conventional training is easy to fall into a local optimal solution; the invention provides a method for combining a GA genetic model and a BP neural network, which solves the network parameters of the BP neural network by combining the GA genetic model with the physical characteristics of an air-conditioning water system and establishes the GA-BP network; in order to further optimize network parameters, optimization training is carried out through a training set and a testing set, an overfitting phenomenon in the training process is prevented through a verification set, and accurate training of the GA-BP network is completed.
Further, the step S11 includes the following sub-steps:
s11-1, setting the hierarchy of a three-layer BP neural network, namely an input layer, a hidden layer and an output layer in sequence;
s11-2, setting the value of the input layer neuron number r as N +1 and the output layer neuron number S2The value of (A) is N, and N is the total number of branches of an air-conditioning water system;
s11-3, calculating to obtain the number S of hidden layer neurons according to the hidden layer node number calculation formula1The value of (c).
The beneficial effects of the above further scheme are: in the process of establishing the three-layer BP neural network, the physical characteristics of an air-conditioning water system are fully combined, and a machine learning neural network which is not redundant and is matched with the air-conditioning water system is established.
Further, the boundary condition of the three-layer BP neural network and the GA genetic model set in step S32 is a numerical range of all input and output feature data of the three-layer BP neural network and the GA genetic model, and is set as a closed interval [ -1,1 ].
The beneficial effects of the above further scheme are: the boundary of the three-layer BP neural network and the GA genetic model is strictly restricted according to engineering requirements, the occurrence of an overrun error in the machine learning training process is prevented, and the system stability and the operation accuracy of the three-layer BP neural network and the GA genetic model are improved.
Further, the step S15 includes the following sub-steps:
s15-1, calculating the weight and the total threshold of the three-layer BP neural network by the following formula:
s＝r×s1+s1×s2+s1+s2
wherein s is the weight and the total threshold of the three layers of BP neural networks;
s15-2, setting vectors with dimensions equal to the values of the weights and the threshold total number of the three-layer BP neural network, and filling all the weights and the thresholds of the three-layer BP neural network into the vectors to obtain weight threshold coding vectors;
s15-3, taking the weight threshold coding vector as an individual of the GA genetic model population;
s15-4, setting the population scale of the GA genetic model, and setting the fitness function of the GA genetic model as 1/f (-) wherein f (-) is the error square sum function of the three-layer BP neural network;
s15-5, randomly generating an initial value of each individual of the population;
s15-6, performing population inheritance, crossing and variation through a GA genetic model according to a GA crossing formula and a GA variation formula;
s15-7, calculating the fitness value of each individual of the inherited, crossed and mutated population according to the fitness function, and eliminating the individuals with the fitness value lower than the fitness threshold;
s15-8, judging whether only one individual remains in the population, if so, jumping to the step S35-9, otherwise, jumping to the step S35-6;
and S15-9, taking the single finally remaining individual of the population as an optimal weight threshold coding vector, extracting to obtain the optimal weight and threshold of the three-layer BP neural network, and completing the establishment of the GA-BP network.
The beneficial effects of the above further scheme are: compared with the method of directly training the three-layer BP neural network, the method for solving the network parameters of the three-layer BP neural network through the GA genetic model is more accurate and is not easy to fall into a local optimal solution, and in the application of the GA genetic model, the design and optimization of the GA genetic model are required to be carried out by combining the network characteristics of the three-layer BP neural network, so that the compatibility and the accuracy are realized.
Further, the GA cross formula of step S15-6 is:
a′kj＝akj(1-b)+aljb
a′lj＝alj(1-b)+akjb
wherein, akjThe j-position genetic gene of the k-th chromosome of the individual, aljIs the j-position genetic gene, a 'of the l chromosome of the individual'kjIs the j position genetic gene, a 'of the k chromosome of the crossed individual'ljThe j position genetic gene of the I chromosome of the crossed individuals.
The beneficial effects of the above further scheme are: the genetic crossing of individuals in the GA genetic model population is carried out through the GA crossing formula, the possibility is provided for the appearance of excellent individuals, and the genetic crossing method is an important way for the evolution and evolution of individuals in the GA genetic model population by combining biological natural science and mathematical knowledge.
Further, the GA variation formula of step S15-6 is:
Figure BDA0003216793370000051
wherein, a ″)kjThe j-position genetic gene of the k-th chromosome of the mutated individual, amaxIs the upper bound value of the genetic Gene, aminIs the lower bound value of the genetic gene, G is the current iteration number of the population, GmaxFor maximum number of iterations of the population, p1Is a first random number, p2Is a second random number, the second random number p2Is a closed interval [0,1]The random number in (c).
The beneficial effects of the above further scheme are: genetic variation of individuals in the GA genetic model population is carried out through a GA variation formula, possibility is provided for appearance of excellent individuals, and the genetic variation method is an important way for evolution and evolution of individuals in the GA genetic model population by combining biological natural science and mathematical knowledge.
Further, the step S16 includes the following sub-steps:
s16-1, setting a target error value, a learning rate value, a data partitioning function, a transfer function from an input layer to a hidden layer and a transfer function from the hidden layer to an output layer of the GA-BP network;
and S16-2, performing parameter optimization training and testing on the GA-BP network according to the training set and the testing set, detecting the output deviation in the training process of the GA-BP network through the verification set in the training and testing processes, continuing the training and testing if the output deviation continuously rises for less than 6 times, and finishing the training of the GA-BP network if the output deviation continuously rises for 6 times.
The beneficial effects of the above further scheme are: after the GA-BP network is established, the GA-BP network is further optimized to be more accurate, and further the pump valve linkage control of the air-conditioning water system is more accurate. And besides the training set and the test set, the verification set is also set, so that the phenomenon of overfitting of the GA-BP network under the training of limited samples is effectively prevented.
Further, the step S3 includes the following sub-steps:
s31, setting an initial value of the current frequency of the water pump of the air conditioner;
s32, inputting the target flow of each branch and the current frequency of the water pump into a GA-BP network, and solving to obtain the opening of each branch balance valve;
s33; judging whether the opening of each branch balancing valve is smaller than or equal to the maximum opening of the balancing valve, if so, skipping to the step S36, and if not, skipping to the step S34;
s34, judging whether the current frequency of the water pump reaches the lower frequency limit, if so, jumping to the step S36, and if not, jumping to the step S35;
s35, adjusting the current frequency of the water pump of the air conditioner down to 0.01Hz, and jumping to the step S32;
and S36, adjusting the balance valve of each branch of the air-conditioning water system according to the opening of the balance valve of each branch to complete pump-valve linkage control.
Further, the initial value of the water pump current frequency of the air conditioner set in the step S31 is 50 Hz.
The beneficial effects of the above further scheme are: when the current frequency of the water pump is too low, the water pump is easy to surge, so that the working state of the water pump is unstable; and the current frequency of the water pump is too high, the power consumption is increased, and the energy conservation is not facilitated, so that 50Hz which does not additionally increase the load of the water pump on the working condition is used as the initial frequency, the downward traversal optimization is carried out on the algorithm design, and the energy consumption and the stability are both considered.
Drawings
Fig. 1 is a schematic flow chart of an air-conditioning water system pump-valve linkage control method based on a neural network.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate the understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and it will be apparent to those skilled in the art that various changes may be made without departing from the spirit and scope of the invention as defined and defined in the appended claims, and all matters produced by the invention using the inventive concept are protected.
As shown in fig. 1, in an embodiment of the present invention, a neural network-based pump-valve linkage control method for an air-conditioning water system comprises the following steps:
and S1, establishing and training a GA-BP network through the measured data of the pump valve of the air-conditioning water system. The method comprises the following steps:
and S11, establishing a three-layer BP neural network.
In the process of establishing a three-layer Back Propagation (BP) neural network, the physical characteristics of an air-conditioning water system are fully combined, and a non-redundant and non-defective machine learning neural network matched with the air-conditioning water system is established. Therefore, in the present embodiment, step S11 includes the following substeps:
s11-1, setting the hierarchy of a three-layer BP neural network, namely an input layer, a hidden layer and an output layer in sequence;
s11-2, setting the value of the input layer neuron number r as N +1 and the output layer neuron number S2The value of (a) is N, wherein N is the total number of branches of the air-conditioning water system, and in the embodiment, the air-conditioning water system comprises 6 branches, namely N is 6;
s11-3, calculating to obtain the number S of hidden layer neurons according to the hidden layer node number calculation formula1The value of (c).
In this embodiment, the hidden layer node number calculation formula is:
Figure BDA0003216793370000071
wherein C is a constant greater than or equal to 0 and less than or equal to 10, and the value of this embodiment is 3.
And S12, setting boundary conditions of the three-layer BP neural network and the GA genetic model. The boundary condition is a numerical range of all input/output feature data of the three-layer BP neural network and the GA genetic model, and is set to a closed interval [ -1,1] in the present embodiment. The GA Genetic model is a computer model of a Genetic Algorithm (GA). The boundary of the three-layer BP neural network and the GA genetic model is strictly restricted according to engineering requirements, the occurrence of an overrun error in the machine learning training process is prevented, and the system stability and the operation accuracy of the three-layer BP neural network and the GA genetic model are improved.
S13, randomly generating opening values of the balance valves of the branches of the M groups of air-conditioning water systems and water pump current frequencies through a computer, controlling the air-conditioning water systems by using the opening values and the water pump current frequencies, actually measuring to obtain corresponding flow values of the branches of the M air-conditioning water systems, and taking the opening values of the balance valves of the branches of the M groups of air-conditioning water systems, the water pump current frequencies and the flow values of the branches of the air-conditioning water systems as sample sets. To make the neural network training sufficient, 900 samples are used in this embodiment, i.e., M is 900.
And S14, randomly dividing the sample set into a training set, a testing set and a verification set.
In this embodiment, there are 800 samples in the training set, 60 samples in the test set, and 40 in the verification set.
And S15, calculating the optimal weight and threshold of the three-layer BP neural network through the GA genetic model according to the training set, and finishing the establishment of the GA-BP network.
Step S15 includes the following substeps:
s15-1, calculating the weight and the total threshold of the three-layer BP neural network by the following formula:
s＝r×s1+s1×s2+s1+s2
wherein s is the total weight and threshold of the three-layer BP neural network.
S15-2, setting a vector with the dimensionality equal to the total weight and threshold value of the three-layer BP neural network, and filling all the weight and threshold values of the three-layer BP neural network into the vector to obtain a weight threshold value coding vector.
And S15-3, taking the weight threshold encoding vector as an individual of the GA genetic model population.
S15-4, setting the population scale of the GA genetic model, and setting the fitness function of the GA genetic model to be 1/f (-) wherein f (-) is the error square sum function of the three-layer BP neural network.
In the present embodiment, the population size is set to 100, and in addition to the population size, the present embodiment also sets the maximum number of iterations of the GA genetic model to 500.
S15-5, randomly generating an initial value of each individual of the population.
And S15-6, performing population inheritance, crossing and mutation through a GA genetic model according to a GA crossing formula and a GA mutation formula.
The GA crossover formula is:
a′kj＝akj(1-b)+aljb
a′lj＝alj(1-b)+akjb
wherein, akjThe j-position genetic gene of the k-th chromosome of the individual, aljIs the j-position genetic gene, a 'of the l chromosome of the individual'kjThe j position genetic base of the k chromosome of the crossed individualsAccordingly, a'ljThe j position genetic gene of the I chromosome of the crossed individuals.
The genetic crossing of individuals in the GA genetic model population is carried out through the GA crossing formula, the possibility is provided for the appearance of excellent individuals, and the genetic crossing method is an important way for the evolution and evolution of individuals in the GA genetic model population by combining biological natural science and mathematical knowledge.
The GA variation formula is as follows:
Figure BDA0003216793370000091
wherein, a ″)kjThe j-position genetic gene of the k-th chromosome of the mutated individual, amaxIs the upper bound value of the genetic Gene, aminIs the lower bound value of the genetic gene, G is the current iteration number of the population, GmaxFor maximum number of iterations of the population, p1Is a first random number, p2Is a second random number, the second random number p2Is a closed interval [0,1]The random number in (c).
The beneficial effects of the above further scheme are: genetic variation of individuals in the GA genetic model population is carried out through a GA variation formula, possibility is provided for appearance of excellent individuals, and the genetic variation method is an important way for evolution and evolution of individuals in the GA genetic model population by combining biological natural science and mathematical knowledge.
S15-7, calculating the fitness value of each individual of the population after inheritance, intersection and variation according to the fitness function, and eliminating the individuals with the fitness value lower than the fitness threshold.
S15-8, judging whether only one individual remains in the population, if so, jumping to the step S35-9, otherwise, jumping to the step S35-6.
And S15-9, taking the single finally remaining individual of the population as an optimal weight threshold coding vector, extracting to obtain the optimal weight and threshold of the three-layer BP neural network, and completing the establishment of the GA-BP network.
Compared with the method of directly training the three-layer BP neural network, the method for solving the network parameters of the three-layer BP neural network through the GA genetic model is more accurate and is not easy to fall into a local optimal solution, and in the application of the GA genetic model, the design and optimization of the GA genetic model are required to be carried out by combining the network characteristics of the three-layer BP neural network, so that the compatibility and the accuracy are realized.
And S16, performing iterative optimization and test verification on the GA-BP network according to the training set, the test set and the verification set to finish the training of the GA-BP network.
The BP neural network is a forward network based on error back propagation, has very strong nonlinear mapping capability, is suitable for solving the opening degree of each branch balance valve of an air-conditioning water system, has the efficiency far higher than that of a traditional air-conditioning water system control method based on a control equation, but the accuracy of the BP neural network depends on a training mode, and the conventional training is easy to fall into a local optimal solution; the invention provides a method for combining a GA genetic model and a BP neural network, which solves the network parameters of the BP neural network by combining the GA genetic model with the physical characteristics of an air-conditioning water system and establishes the GA-BP network; in order to further optimize network parameters, optimization training is carried out through a training set and a testing set, an overfitting phenomenon in the training process is prevented through a verification set, and accurate training of the GA-BP network is completed.
Step S16 includes the following substeps:
s16-1, setting a target error value of the GA-BP network, a learning rate value, a data division function, a transfer function from an input layer to a hidden layer and a transfer function from the hidden layer to an output layer.
In the present embodiment, the target error value is set to 10-5The learning rate value is set to 0.01, and the data partitioning function is a dividerand function provided by Matlab software.
The input layer to hidden layer transfer function is set as:
Figure BDA0003216793370000111
wherein e is a natural constant, and b is an exponential coefficient.
The transfer function from the hidden layer to the output layer is set as purelin function provided by Matlab software.
And S16-2, performing parameter optimization training and testing on the GA-BP network according to the training set and the testing set, detecting the output deviation in the training process of the GA-BP network through the verification set in the training and testing processes, continuing the training and testing if the output deviation continuously rises for less than 6 times, and finishing the training of the GA-BP network if the output deviation continuously rises for 6 times.
After the GA-BP network is established, the GA-BP network is further optimized to be more accurate, and further the pump valve linkage control of the air-conditioning water system is more accurate. And besides the training set and the test set, the verification set is also set, so that the phenomenon of overfitting of the GA-BP network under the training of limited samples is effectively prevented.
And S2, calculating the target flow of each branch required by the air-conditioning water system balance according to the real-time load of the air conditioner.
During the operation and use of the central air conditioner, the load of the central air conditioner changes with the change of factors such as outdoor environment, indoor personnel, heat sources and the like. Under the premise that the temperature difference between the supplied water and the returned water is not changed, when the load of the air conditioner is changed, if the water flow of each user branch of the air conditioner water system is not adjusted in time, the phenomenon of uneven cooling and heating at each user can be caused, and dynamic hydraulic imbalance is caused.
In the embodiment, the hydraulic imbalance condition is detected by a flowmeter arranged on each branch, and the target flow of each branch required by the balance of the air-conditioning water system is the difference value between the average value of the flow of each branch of the air conditioner and the real-time flow value detected by the flowmeter.
And S3, solving through a GA-BP network according to the target flow of each branch to obtain the opening of each branch balance valve and the current frequency of the water pump.
Step S3 includes the following substeps:
and S31, setting an initial value of the current frequency of the water pump of the air conditioner.
In this embodiment, the initial value of the water pump current frequency of the air conditioner is set to 50 Hz.
When the current frequency of the water pump is too low, the water pump is easy to surge, so that the working state of the water pump is unstable; and the current frequency of the water pump is too high, the power consumption is increased, and the energy conservation is not facilitated, so that 50Hz which does not additionally increase the load of the water pump on the working condition is used as the initial frequency, the downward traversal optimization is carried out on the algorithm design, and the energy consumption and the stability are both considered.
And S32, inputting the target flow of each branch and the current frequency of the water pump into the GA-BP network, and solving to obtain the opening of each branch balance valve.
S33; and judging whether the opening of each branch balance valve is smaller than or equal to the maximum opening of the balance valve, if so, jumping to the step S36, and if not, jumping to the step S34.
And S34, judging whether the current frequency of the water pump reaches the lower frequency limit, if so, jumping to the step S36, and if not, jumping to the step S35.
S35, adjusting the current frequency of the water pump of the air conditioner down to 0.01Hz, and jumping to the step S32.
And S36, adjusting the balance valve of each branch of the air-conditioning water system according to the opening of the balance valve of each branch to complete pump-valve linkage control.
And S4, adjusting the balance valve of each branch of the air-conditioning water system according to the opening of the balance valve of each branch, and controlling the water pump according to the current frequency of the water pump to complete the linkage control of the pump and the valve.
The key of the accuracy of the pump-valve linkage control lies in the accuracy of the calculation of the GA-BP network, and in order to verify the advancement of the GA-BP network of the invention relative to the prior art, the embodiment is provided with a contrast experiment.
TABLE 1 comparison of the Performance of GA-BP networks of the present invention embodiments with conventional BP neural networks
Figure BDA0003216793370000121
Figure BDA0003216793370000131
As can be seen from Table 1, the mean square error of the GA-BP network of the present invention is significantly smaller than that of the conventional BP neural network.
Besides the effect of the GA-BP network, the pump-valve linkage actual control effect of the method of the invention also needs to be tested. In the embodiment, the pump valve linkage control effect of the air-conditioning water system is tested in 10 different working condition environments, and the test results are shown in table 2:
table 2 pump valve linkage control effect comparison table under 10 different working condition environments according to the embodiment of the present invention
Figure BDA0003216793370000132
Figure BDA0003216793370000141
The relative errors of the pump valve linkage control of the air-conditioning water system under each working condition in the table 2 are counted, and the result is shown in the table 3:
TABLE 3 Total flow relative error of single branch loop 10 group operation condition system
Figure BDA0003216793370000142
As can be seen from table 3, through actual measurement in this embodiment, the total flow error of the control effect of the pump-valve linkage control method for the air-conditioning water system based on the neural network provided by the invention is not greater than 3.6%, and the control effect is accurate.
In conclusion, the GA-BP network combining the GA genetic model and the BP neural network is used as the operation core, the opening degree of each branch balance valve can be rapidly solved under the condition that the target flow of each branch of the air-conditioning water system and the current frequency of the water pump are known, and the problems of slow response and inaccurate adjustment of the pump-valve linkage in the prior art are solved; the unidirectional water pump current frequency traversal optimization searching method also has the advantages of rapidness and effectiveness; benefit from the quick accurate operation of pump valve linkage parameter, the pump valve linkage need not retrace repeatedly and educates and produce unnecessary consumption, through the settlement of appropriate water pump current frequency, compares prior art but the whole consumption of greatly reduced.
The principle and the implementation mode of the invention are explained by applying specific embodiments in the invention, and the description of the embodiments is only used for helping to understand the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
It will be appreciated by those of ordinary skill in the art that the embodiments described herein are intended to assist the reader in understanding the principles of the invention and are to be construed as being without limitation to such specifically recited embodiments and examples. Those skilled in the art can make various other specific changes and combinations based on the teachings of the present invention without departing from the spirit of the invention, and these changes and combinations are within the scope of the invention.

Claims (10)

1. A pump and valve linkage control method of an air conditioning water system based on a neural network is characterized by comprising the following steps:
s1, establishing and training a GA-BP network through actually measured data of a pump valve of an air-conditioning water system;
s2, calculating target flow of each branch required by air conditioner water system balance according to real-time load of the air conditioner;
s3, solving through a GA-BP network according to the target flow of each branch to obtain the opening of each branch balance valve and the current frequency of the water pump;
and S4, adjusting the balance valve of each branch of the air-conditioning water system according to the opening of the balance valve of each branch, and controlling the water pump according to the current frequency of the water pump to complete the linkage control of the pump and the valve.
2. An air conditioning water system pump valve linkage control method based on a neural network as claimed in claim 1, wherein the step S1 includes the following substeps:
s11, establishing a three-layer BP neural network;
s12, setting boundary conditions of a three-layer BP neural network and a GA genetic model;
s13, randomly generating opening values of all branch balance valves and water pump current frequencies of M groups of air-conditioning water systems through a computer, controlling the air-conditioning water systems by using the opening values and the water pump current frequencies, actually measuring to obtain corresponding flow values of all branches of the M air-conditioning water systems, and taking the opening values of all branch balance valves, the water pump current frequencies and the flow values of all branches of the air-conditioning water systems of the M groups as sample sets;
s14, randomly dividing the sample set into a training set, a testing set and a verification set;
s15, according to the training set, calculating the optimal weight and threshold of the three-layer BP neural network through a GA genetic model, and completing the establishment of the GA-BP network;
and S16, performing iterative optimization and test verification on the GA-BP network according to the training set, the test set and the verification set to finish the training of the GA-BP network.
3. An air conditioning water system pump valve linkage control method based on a neural network as claimed in claim 2, wherein the step S11 includes the following substeps:
s11-1, setting the hierarchy of a three-layer BP neural network, namely an input layer, a hidden layer and an output layer in sequence;
s11-2, setting the value of the input layer neuron number r as N +1 and the output layer neuron number S2The value of (A) is N, and N is the total number of branches of an air-conditioning water system;
s11-3, calculating to obtain the number S of hidden layer neurons according to the hidden layer node number calculation formula1The value of (c).
4. An air-conditioning water system pump valve linkage control method based on a neural network as claimed in claim 3, wherein the boundary conditions of the three-layer BP neural network and the GA genetic model set in the step S12 are numerical ranges of all input and output characteristic data of the three-layer BP neural network and the GA genetic model, and are set as a closed interval [ -1,1 ].
5. An air conditioning water system pump valve linkage control method based on a neural network as claimed in claim 4, wherein the step S15 includes the following substeps:
s15-1, calculating the weight and the total threshold of the three-layer BP neural network by the following formula:
s＝r×s1+s1×s2+s1+s2
wherein s is the weight and the total threshold of the three layers of BP neural networks;
s15-2, setting vectors with dimensions equal to the values of the weights and the threshold total number of the three-layer BP neural network, and filling all the weights and the thresholds of the three-layer BP neural network into the vectors to obtain weight threshold coding vectors;
s15-3, taking the weight threshold coding vector as an individual of the GA genetic model population;
s15-4, setting the population scale of the GA genetic model, and setting the fitness function of the GA genetic model as 1/f (-) wherein f (-) is the error square sum function of the three-layer BP neural network;
s15-5, randomly generating an initial value of each individual of the population;
s15-6, performing population inheritance, crossing and variation through a GA genetic model according to a GA crossing formula and a GA variation formula;
s15-7, calculating the fitness value of each individual of the inherited, crossed and mutated population according to the fitness function, and eliminating the individuals with the fitness value lower than the fitness threshold;
s15-8, judging whether only one individual remains in the population, if so, jumping to the step S15-9, otherwise, jumping to the step S15-6;
and S15-9, taking the single finally remaining individual of the population as an optimal weight threshold coding vector, extracting to obtain the optimal weight and threshold of the three-layer BP neural network, and completing the establishment of the GA-BP network.
6. An air-conditioning water system pump valve linkage control method based on a neural network as claimed in claim 5, wherein the GA cross formula of step S15-6 is:
a′kj＝akj(1-b)+aljb
a′lj＝alj(1-b)+akjb
wherein, akjThe j-position genetic gene of the k-th chromosome of the individual, aljIs the j-position genetic gene, a 'of the l chromosome of the individual'kjIs the j position genetic gene, a 'of the k chromosome of the crossed individual'ljThe j position genetic gene of the I chromosome of the crossed individuals.
7. An air-conditioning water system pump valve linkage control method based on a neural network as claimed in claim 6, wherein the GA variation formula of step S15-6 is:
Figure FDA0003216793360000031
wherein, a ″)kjThe j-position genetic gene of the k-th chromosome of the mutated individual, amaxIs the upper bound value of the genetic Gene, aminIs the lower bound value of the genetic gene, G is the current iteration number of the population, GmaxFor maximum number of iterations of the population, p1Is a first random number, p2Is a second random number, the second random number p2Is a closed interval [0,1]The random number in (c).
8. An air conditioning water system pump valve linkage control method based on a neural network as claimed in claim 7, wherein the step S16 includes the following substeps:
s16-1, setting a target error value, a learning rate value, a data partitioning function, a transfer function from an input layer to a hidden layer and a transfer function from the hidden layer to an output layer of the GA-BP network;
and S16-2, performing parameter optimization training and testing on the GA-BP network according to the training set and the testing set, detecting the output deviation in the training process of the GA-BP network through the verification set in the training and testing processes, continuing the training and testing if the output deviation continuously rises for less than 6 times, and finishing the training of the GA-BP network if the output deviation continuously rises for 6 times.
9. An air conditioning water system pump valve linkage control method based on a neural network as claimed in claim 8, wherein the step S3 includes the following substeps:
s31, setting an initial value of the current frequency of the water pump of the air conditioner;
s32, inputting the target flow of each branch and the current frequency of the water pump into a GA-BP network, and solving to obtain the opening of each branch balance valve;
s33; judging whether the opening of each branch balancing valve is smaller than or equal to the maximum opening of the balancing valve, if so, skipping to the step S36, and if not, skipping to the step S34;
s34, judging whether the current frequency of the water pump reaches the lower frequency limit, if so, jumping to the step S36, and if not, jumping to the step S35;
s35, adjusting the current frequency of the water pump of the air conditioner down to 0.01Hz, and jumping to the step S32;
and S36, adjusting the balance valve of each branch of the air-conditioning water system according to the opening of the balance valve of each branch to complete pump-valve linkage control.
10. An air-conditioning water system pump valve linkage control method based on a neural network as claimed in claim 9, wherein the initial value of the water pump current frequency of the air conditioner set in the step S31 is 50 Hz.
CN202110946317.4A 2021-08-18 2021-08-18 Air-conditioning water system pump valve linkage control method based on neural network Active CN113654143B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110946317.4A CN113654143B (en) 2021-08-18 2021-08-18 Air-conditioning water system pump valve linkage control method based on neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110946317.4A CN113654143B (en) 2021-08-18 2021-08-18 Air-conditioning water system pump valve linkage control method based on neural network

Publications (2)

Publication Number Publication Date
CN113654143A true CN113654143A (en) 2021-11-16
CN113654143B CN113654143B (en) 2023-04-07

Family

ID=78480755

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110946317.4A Active CN113654143B (en) 2021-08-18 2021-08-18 Air-conditioning water system pump valve linkage control method based on neural network

Country Status (1)

Country Link
CN (1) CN113654143B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115671328A (en) * 2022-11-08 2023-02-03 上海博迅医疗生物仪器股份有限公司 A sterilization system for culture medium preparation

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2003207189A (en) * 2002-01-18 2003-07-25 Foundation For The Promotion Of Industrial Science Indoor thermal environment design system and its method
CN101498534A (en) * 2008-12-08 2009-08-05 天津大学 Multi-target intelligent control method for electronic expansion valve of refrigeration air conditioner heat pump system
CN101769586A (en) * 2010-02-04 2010-07-07 无锡永信能源科技有限公司 Cold (warm) water circulation energy efficiency control method and cold (warm) water circulation energy efficiency control device for central air-conditioning system
CN106765959A (en) * 2016-12-27 2017-05-31 武汉虹信技术服务有限责任公司 Heat-air conditioner energy-saving control method based on genetic algorithm and depth B P neural network algorithms
CN111898260A (en) * 2020-07-20 2020-11-06 四川省建筑科学研究院有限公司 Variable flow optimization control method and controller for central air conditioning system

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2003207189A (en) * 2002-01-18 2003-07-25 Foundation For The Promotion Of Industrial Science Indoor thermal environment design system and its method
CN101498534A (en) * 2008-12-08 2009-08-05 天津大学 Multi-target intelligent control method for electronic expansion valve of refrigeration air conditioner heat pump system
CN101769586A (en) * 2010-02-04 2010-07-07 无锡永信能源科技有限公司 Cold (warm) water circulation energy efficiency control method and cold (warm) water circulation energy efficiency control device for central air-conditioning system
CN106765959A (en) * 2016-12-27 2017-05-31 武汉虹信技术服务有限责任公司 Heat-air conditioner energy-saving control method based on genetic algorithm and depth B P neural network algorithms
CN111898260A (en) * 2020-07-20 2020-11-06 四川省建筑科学研究院有限公司 Variable flow optimization control method and controller for central air conditioning system

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115671328A (en) * 2022-11-08 2023-02-03 上海博迅医疗生物仪器股份有限公司 A sterilization system for culture medium preparation
CN115671328B (en) * 2022-11-08 2023-12-05 上海博迅医疗生物仪器股份有限公司 Sterilization system for preparing culture medium

Also Published As

Publication number Publication date
CN113654143B (en) 2023-04-07

Similar Documents

Publication Publication Date Title
CN106920006B (en) Subway station air conditioning system energy consumption prediction method based on ISOA-LSSVM
CN108694467B (en) Method and system for predicting line loss rate of power distribution network
CN109084415A (en) Central air-conditioning operating parameter optimization method based on artificial neural network and genetic algorithms
CN107316099A (en) Ammunition Storage Reliability Forecasting Methodology based on particle group optimizing BP neural network
CN107704875A (en) Based on the building load Forecasting Methodology and device for improving IHCMAC neutral nets
CN104636985A (en) Method for predicting radio disturbance of electric transmission line by using improved BP (back propagation) neural network
CN103105246A (en) Greenhouse environment forecasting feedback method of back propagation (BP) neural network based on improvement of genetic algorithm
CN108564205A (en) A kind of load model and parameter identification optimization method based on measured data
CN105608295B (en) The multi-objective genetic algorithm of coking furnace pressure and RBF neural Optimization Modeling method
CN109376933A (en) Lithium ion battery negative material energy density prediction technique neural network based
CN113654143B (en) Air-conditioning water system pump valve linkage control method based on neural network
CN113972667A (en) Power distribution network reactive power optimization method based on one-dimensional convolutional neural network
CN108038797B (en) Two-term coefficient method and multi-target planning hybrid evaluation method for power distribution network scheduling control level
CN106875055A (en) Energy storage device based on AHP and GA participates in the parameter optimization method of power grid frequency modulation
CN114216256A (en) Ventilation system air volume control method of off-line pre-training-on-line learning
CN115860170A (en) Power quality optimization method of power distribution system considering power electronic load
CN114936742A (en) Water supply system scheduling agent decision method
Shang et al. Research on intelligent pest prediction of based on improved artificial neural network
CN105633971B (en) A kind of intelligent distribution network probabilistic loadflow method based on imperial competition algorithm
CN110298506A (en) A kind of urban construction horizontal forecast system
CN110033118A (en) Elastomeric network modeling and the blower multiobjective optimization control method based on genetic algorithm
CN111291465A (en) Ship air conditioner thermal comfort temperature prediction method based on genetic algorithm and neural network
CN110210623A (en) Adaptive multiple target hybrid differential evolution algorithm based on simulated annealing and comentropy
CN113642784B (en) Wind power ultra-short-term prediction method considering fan state
CN109086952A (en) It is a kind of based on genetic algorithm-neural network heat load prediction method

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