CN114046259B - Centrifugal pump variable frequency control method based on double neural network model - Google Patents
Centrifugal pump variable frequency control method based on double neural network model Download PDFInfo
- Publication number
- CN114046259B CN114046259B CN202111340207.XA CN202111340207A CN114046259B CN 114046259 B CN114046259 B CN 114046259B CN 202111340207 A CN202111340207 A CN 202111340207A CN 114046259 B CN114046259 B CN 114046259B
- Authority
- CN
- China
- Prior art keywords
- pressure
- flow
- neural network
- est
- centrifugal pump
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 24
- 238000003062 neural network model Methods 0.000 title claims abstract description 17
- 238000013528 artificial neural network Methods 0.000 claims abstract description 58
- 230000001105 regulatory effect Effects 0.000 claims abstract description 38
- 230000033228 biological regulation Effects 0.000 claims abstract description 11
- 238000012549 training Methods 0.000 claims description 23
- 210000002569 neuron Anatomy 0.000 claims description 22
- 230000009977 dual effect Effects 0.000 claims description 10
- 238000011056 performance test Methods 0.000 claims description 9
- 238000012360 testing method Methods 0.000 claims description 9
- 238000012795 verification Methods 0.000 claims description 6
- 238000010219 correlation analysis Methods 0.000 claims description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 abstract description 24
- 238000005265 energy consumption Methods 0.000 abstract description 3
- 238000012544 monitoring process Methods 0.000 abstract description 3
- 238000006243 chemical reaction Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 6
- 230000006641 stabilisation Effects 0.000 description 5
- 238000011105 stabilization Methods 0.000 description 5
- 230000001276 controlling effect Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011217 control strategy Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F04—POSITIVE - DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS FOR LIQUIDS OR ELASTIC FLUIDS
- F04D—NON-POSITIVE-DISPLACEMENT PUMPS
- F04D15/00—Control, e.g. regulation, of pumps, pumping installations or systems
- F04D15/0066—Control, e.g. regulation, of pumps, pumping installations or systems by changing the speed, e.g. of the driving engine
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F04—POSITIVE - DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS FOR LIQUIDS OR ELASTIC FLUIDS
- F04D—NON-POSITIVE-DISPLACEMENT PUMPS
- F04D15/00—Control, e.g. regulation, of pumps, pumping installations or systems
- F04D15/0005—Control, e.g. regulation, of pumps, pumping installations or systems by using valves
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F04—POSITIVE - DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS FOR LIQUIDS OR ELASTIC FLUIDS
- F04D—NON-POSITIVE-DISPLACEMENT PUMPS
- F04D15/00—Control, e.g. regulation, of pumps, pumping installations or systems
- F04D15/0088—Testing machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Mechanical Engineering (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Control Of Non-Positive-Displacement Pumps (AREA)
Abstract
The invention belongs to the technical field of water pumps, and particularly relates to a centrifugal pump variable frequency control method based on a double-neural network model, wherein the input power P act and the working speed frequency f act of a motor are used as the input of a controller, the current working pressure P est and the current working flow q est are obtained by means of a double-neural network prediction model, the controller outputs the regulating speed frequency f o based on the deviation between the current working pressure P est and a required working pressure value P set, the regulating speed frequency f o is used as the input frequency of a frequency converter, and the working speeds of the motor and the centrifugal pump are controlled; based on the deviation of the current working flow q est from the required working flow value q set, the controller outputs a control signal to adjust the opening of the electric control valveAdjusting the opening of an electric control valveAs an input value to the electric control valve actuator, the opening degree of the electric control valve is controlled. The invention does not need a pressure sensor and a flow sensor, realizes the real-time monitoring and intelligent regulation of the running condition of the centrifugal pump under the condition of no pressure sensor and flow sensor, and has high control reliability and low energy consumption.
Description
Technical Field
The invention belongs to the technical field of water pumps, and particularly relates to a centrifugal pump variable frequency control method based on a double neural network model.
Background
The traditional variable frequency control method of the centrifugal pump mostly realizes accurate control of the output pressure of the centrifugal pump by means of the feedback signal of the pressure sensor.
The Chinese patent CN208650145U discloses a constant-pressure water supply system, which comprises a frequency converter, a water pump motor, a first pressure switch for detecting the upper limit of water pressure and a second pressure switch for detecting the lower limit of water pressure, wherein the power end of the frequency converter is connected with a power grid, the function selection input end of the frequency converter is connected with the switch signal output ends of the first pressure switch and the second pressure switch, and the output control end of the frequency converter is connected with the water pump motor. The variable frequency control is realized through the first pressure switch, the second pressure switch and the frequency converter, so that constant pressure water supply is realized.
The utility model patent CN208533622U discloses a frequency conversion constant pressure water supply device, which comprises a steady flow water tank, a pump group, a PLC controller and a frequency converter, wherein a pressure transmitter is arranged at the water outlet end of the pump group, the pump group comprises a plurality of frequency conversion water supply branches which are connected in parallel and are provided with frequency conversion pumps, the PLC controller is respectively and electrically connected with the pressure transmitter and the frequency converter, the frequency converter controls the frequency conversion operation of the frequency conversion pumps, the pump group also comprises a pressure stabilization water supply branch which is connected with any frequency conversion water supply branch in parallel and is provided with a pressure stabilization pump, the pressure stabilization pump is connected with the PLC controller, the output power of the pressure stabilization pump is smaller than the lowest frequency output power of the frequency conversion pumps, a circulating pipeline which is communicated with the steady flow water tank is arranged at the water outlet end of the pressure stabilization water supply branch, a water purifier and an electric butterfly valve are arranged on the circulating pipeline, and the electric butterfly valve is connected with the PLC controller. The variable frequency control of the water pump is realized by adopting a PLC, a frequency converter and a pressure transmitter and combining a PID control algorithm, so that the constant outlet pressure is ensured.
However, due to physical space limitations or cost-based considerations, a corresponding pressure sensor cannot be configured for the centrifugal pump assembly, which would result in an inability to achieve accurate control of the centrifugal pump operating conditions.
Disclosure of Invention
The invention aims to provide a centrifugal pump variable frequency control method based on a double neural network model, which does not need a pressure sensor and a flow sensor, realizes real-time monitoring and intelligent regulation of the running condition of the centrifugal pump under the condition of no pressure sensor and flow sensor, and has high control reliability and low energy consumption.
The purpose of the invention is realized in the following way:
a centrifugal pump variable frequency control method based on a double neural network model comprises the following steps:
a1, setting a working pressure value p set and a pressure allowable error delta p required by the centrifugal pump;
A2, measuring the input power P act and the working rotation speed frequency f act of the motor, and obtaining the current working pressure P est of the centrifugal pump by means of a double-neural network prediction model;
A3, based on the deviation between the current working pressure p est and the required working pressure value p set, the controller outputs the regulated rotating speed frequency f o, the regulated rotating speed frequency f o is used as the input frequency of the frequency converter, and the working rotating speeds of the motor and the centrifugal pump are controlled;
A4, judging whether the absolute value of the deviation between the current working pressure p est and the required working pressure p set is smaller than a pressure allowable error delta p, if so, entering a step A5; if not, returning to the step A2, and performing iteration control;
A5, finishing regulation;
Or comprises the following steps:
B1, setting a required working pressure value p set, a required working flow value q set, a pressure allowable error delta p, a flow allowable error delta q and an initial opening of an electric regulating valve of the centrifugal pump;
B2, measuring the input power P act and the working rotation speed frequency f act of the motor, and obtaining the current working pressure P est and the current working flow q est of the centrifugal pump by means of a double-neural network prediction model;
b3, based on the deviation between the current working pressure p est and the required working pressure value p set, the controller outputs the regulated rotating speed frequency f o, the regulated rotating speed frequency f o is used as the input frequency of the frequency converter, and the working rotating speeds of the motor and the centrifugal pump are controlled;
b4, judging whether the absolute value of the deviation between the current working pressure p est and the required working pressure p set is smaller than a pressure allowable error delta p, if so, entering a step B5; if not, returning to the step B2, and performing iteration control;
B5, based on the deviation of the current working flow q est and the required working flow value q set, the controller outputs and adjusts the opening of the electric regulating valve Will adjust the electric control valve opening/>As an input value of the actuating mechanism, controlling the opening of the electric regulating valve;
B6, judging whether the absolute value of the deviation of the current working pressure p est and the required working pressure p set is smaller than a pressure allowable error delta p, judging whether the absolute value of the deviation of the current working flow q est and the required working flow q set is smaller than a flow allowable error delta q, and if the absolute value of the deviation of the current working pressure p est and the required working pressure p set are both satisfied, entering a step B7; if at least one item does not meet, returning to the step B2, and performing iterative control;
B7, finishing regulation;
The dual-neural network prediction model adopts a three-layer counter-propagation BP neural network and comprises a flow neural network prediction model NN1 and a pressure neural network prediction model NN2, wherein an input layer of the flow neural network prediction model NN1 comprises 2 neurons which are respectively a motor input power P act and a working rotating speed frequency f act, a hidden layer of the flow neural network prediction model NN1 comprises a plurality of neurons, and an output layer of the flow neural network prediction model NN1 comprises 1 neuron and is a current working flow q est; the input layer of the pressure neural network prediction model NN2 comprises 2 neurons, namely a working rotation speed frequency f act and a current working flow q est,, the hidden layer of the pressure neural network prediction model NN2 comprises a plurality of neurons, and the output layer of the pressure neural network prediction model NN2 comprises 1 neuron and is a current working pressure p est.
In the centrifugal pump variable frequency control method based on the dual neural network model, a centrifugal pump hydraulic performance test dataset is obtained, wherein the centrifugal pump hydraulic performance test dataset comprises a current working pressure P est, a current working flow q est, a motor input power P act and a working rotating speed frequency f act, the centrifugal pump hydraulic performance test dataset is based on the centrifugal pump hydraulic performance test dataset, a flow neural network prediction model NN1 and a pressure neural network prediction model NN2 are trained respectively, and test data are distributed into a training sample dataset, a verification sample dataset and a test sample dataset according to the proportion of 80%, 10% and 10% by adopting a random distribution method.
In the centrifugal pump variable frequency control method based on the dual neural network model, the training algorithm adopts a Bayesian regularization algorithm, the training iteration number is set to 5000, the learning rate is set to 0.1, the target error is set to 0.00001, and after the training is completed, correlation analysis between target values and output values is carried out on the training result, the verification result and the test result respectively.
In the centrifugal pump variable frequency control method based on the dual neural network model, the hidden layer of the flow neural network prediction model NN1 comprises 10 neurons, and the hidden layer of the pressure neural network prediction model NN2 comprises 10 neurons, so that the operation time is reduced while the output accuracy is ensured.
In the centrifugal pump variable frequency control method based on the dual neural network model, the controller comprises a pressure PI controller and a flow PI controller, the pressure PI controller outputs a rotation speed frequency f o to control the working rotation speeds of the motor and the centrifugal pump; output of flow PI controller adjusts opening of electric regulating valveAnd controlling the opening degree of the electric regulating valve.
In the above-mentioned centrifugal pump frequency conversion control method based on the dual neural network model, the motor input power P act and the working rotation speed frequency f act in step A2 or step B2 are read by a frequency converter or measured by a power sensor and a hall sensor connected to the motor.
Compared with the prior art, the invention has the following outstanding and beneficial technical effects:
The invention realizes the accurate control of the operation condition of the centrifugal pump under the condition of no need of a pressure sensor and a flow sensor by means of the dual-neural network model and the controller, reduces the manufacturing cost and the maintenance cost of equipment, monitors and intelligently and accurately regulates the centrifugal pump in real time under the condition of no pressure sensor and flow sensor by means of the dual-neural network model, and ensures the operation safety and reliability of the centrifugal pump.
Drawings
FIG. 1 is a flow chart of a control method according to a second embodiment of the present invention;
FIG. 2 is a flow chart of a control method according to a first embodiment of the present invention;
FIG. 3 is a schematic diagram of a water supply system according to a second embodiment of the present invention;
FIG. 4 is a schematic view of a water supply system according to a first embodiment of the present invention;
FIG. 5 is a schematic diagram of a control system according to a second embodiment of the present invention;
FIG. 6 is a schematic diagram of a control system according to a first embodiment of the present invention;
FIG. 7 is a schematic diagram of a dual neural network prediction model of the present invention;
FIG. 8 is a schematic diagram of the flow-pressure performance curve of the present invention;
fig. 9 is a schematic diagram of the flow-power performance curve of the present invention.
Reference numerals: 1. a water tank; 2. a centrifugal pump; 3. a motor; 4. a frequency converter; 5. a controller; 6. an electric control valve; 7. and a hand valve.
Detailed Description
The invention is further described in the following examples with reference to the accompanying drawings:
Embodiment one: see fig. 2, 4, 6, 7, 8, 9:
A variable frequency centrifugal pump with rated flow rate of q n＝15m3/h, rated pressure of P n =700 kPa, rated rotating speed of n=2900r/min, rated rotating speed frequency of f n =48.33 Hz and rated power of P n =5.5 kW is taken as an analysis object,
A centrifugal pump variable frequency control method based on a double neural network model comprises the following steps:
B1, setting a required working pressure value p set of the centrifugal pump 2 to be 500kPa, a required working flow value q set to be 10m 3/h, a pressure allowable error delta p to be 3kPa, a flow allowable error delta q to be 0.5m 3/h and an initial opening of an electric regulating valve to be 50%;
B2, measuring the input power P act and the working rotation speed frequency f act of the motor, and obtaining the current working pressure P est and the current working flow q est of the centrifugal pump 2 by means of a double-neural network prediction model;
B3, based on the deviation between the current working pressure p est and the required working pressure value p set, the pressure PI controller outputs the regulated rotating speed frequency f o through proportional integral operation, the regulated rotating speed frequency f o is used as the input frequency of the frequency converter 4, the working rotating speeds of the motor 3 and the centrifugal pump 2 are controlled, and the output pressure of the centrifugal pump 2 is controlled;
b4, judging whether the absolute value of the deviation between the current working pressure p est and the required working pressure p set is smaller than a pressure allowable error delta p, if so, entering a step B5; if not, returning to the step B2, and performing iteration control;
B5, based on the deviation between the current working flow q est and the required working flow q set, the flow PI controller outputs and adjusts the opening of the electric regulating valve through proportional integral operation Will adjust the electric control valve opening/>As an input value of an executing mechanism of the electric regulating valve 6, controlling the opening degree of the electric regulating valve 6 to realize the control of the output flow of the centrifugal pump 2;
B6, judging whether the absolute value of the deviation of the current working pressure p est and the required working pressure p set is smaller than a pressure allowable error delta p, judging whether the absolute value of the deviation of the current working flow q est and the required working flow q set is smaller than a flow allowable error delta q, and if the absolute value of the deviation of the current working pressure p est and the required working pressure p set are both satisfied, entering a step B7; if at least one item does not meet, returning to the step B2, and performing iterative control;
B7, finishing regulation;
The required working flow value q set in the step B1 is set according to the water consumption condition of the user, that is, the opening condition of the hand valve 7, and the more the hand valve 7 is opened, the larger the required working flow is, the larger the required working flow value q set is, otherwise, the smaller the required working flow is, and the smaller the required working flow value q set is.
In step B2, the motor input power P act and the working rotation speed frequency f act are read by a frequency converter or measured by a power sensor and a hall sensor connected to the motor, the power sensor measures the motor input power P act, the hall sensor measures the working rotation speed frequency f act, the motor input power P act and the working rotation speed frequency f act are used as the input of the controller 5, and the controller 5 outputs the regulating rotation speed frequency f o and the opening degree of the electric regulating valve by means of a dual neural network prediction model and a PI control algorithm
As shown in fig. 4, the controller 5 outputs the rotation speed adjusting frequency f o to the frequency converter 4, the frequency converter 4 controls the working rotation speed of the motor 3, the motor 3 drives the centrifugal pump 3 to work, and the greater the rotation speed of the motor 3 is, the greater the working pressure of the centrifugal pump 2 is; the controller 5 outputs and adjusts the opening of the electric regulating valveThe larger the opening of the electric control valve 6 is, the larger the operating flow rate of the centrifugal pump 2 is, to the electric control valve 6.
As shown in fig. 7, the dual-neural network prediction model adopts a three-layer back propagation BP neural network, including a flow neural network prediction model NN1 and a pressure neural network prediction model NN2, where an input layer of the flow neural network prediction model NN1 includes 2 neurons, which are respectively a motor input power P act and a working rotation speed frequency f act, a hidden layer of the flow neural network prediction model NN1 includes 10 neurons, and an output layer of the flow neural network prediction model NN1 includes 1 neuron, which is a current working flow q est; the input layer of the pressure neural network prediction model NN2 contains 2 neurons, namely the working rotation speed frequency f act and the current working flow q est, the hidden layer of the pressure neural network prediction model NN2 contains 10 neurons, and the output layer of the pressure neural network prediction model NN2 contains 1 neuron, which is the current working pressure p est.
The working principle of the embodiment is as follows: as shown in fig. 2, 4 and 6, the input power P act and the working rotation speed frequency f act of the motor are used as the input of a flow neural network prediction model NN1 in a dual neural network prediction model, the current working flow q est and the working rotation speed frequency f act output by the flow neural network prediction model NN1 are used as the input of a pressure neural network prediction model NN2, the pressure neural network prediction model NN2 outputs the current working pressure P est, the pressure PI controller outputs the regulating rotation speed frequency f o based on the deviation between the current working pressure P est and the required working pressure value P set, the regulating rotation speed frequency f o is used as the input frequency of the frequency converter 4, and the working rotation speeds of the motor 3 and the centrifugal pump 2 are controlled until the deviation between the current working pressure P est and the required working pressure value P set is smaller than the pressure allowable error delta p, so that the working pressure of the centrifugal pump 2 is always within a certain range, and the accurate control of the pressure of the centrifugal pump 2 is realized; based on the deviation of the current working flow q est and the required working flow value q set, the flow PI controller outputs and adjusts the opening of the electric regulating valveAdjusting the opening of the electric adjusting valve/>As the input value of the actuating mechanism of the electric regulating valve 6, the opening of the electric regulating valve 6 is controlled until the deviation between the current working flow q est and the required working flow q set is smaller than the flow allowable error delta q, so that the working flow of the centrifugal pump 2 is always in a certain range, the accurate control of the flow of the centrifugal pump 2 is realized, the real-time monitoring and intelligent accurate regulation and control are realized under the condition of no pressure sensor or flow sensor, the reliability of control is high, and the energy consumption is low.
A hydraulic performance test dataset of the centrifugal pump 2 is obtained, comprising a current working pressure P est, a current working flow q est, a motor input power P act and a working rotation speed frequency f act, and a flow-pressure sample curve (shown in fig. 8) and a flow-power sample curve (shown in fig. 9) of the centrifugal pump 2 are established. Based on the hydraulic performance test data set of the centrifugal pump 2, the flow neural network prediction model NN1 and the pressure neural network prediction model NN2 are trained respectively, and the test data are distributed into a training sample data set, a verification sample data set and a test sample data set according to the proportion of 80%, 10% and 10% by adopting a random distribution method. The algorithm of the dual neural network prediction model adopts a Bayesian regularization algorithm, the iteration number of training is set to 5000, the learning rate is set to 0.1, the target error is set to 0.00001, and after training is completed, correlation analysis between target values and output values is respectively carried out on a training result, a verification result and a test result, namely R value analysis is carried out, and the closer the R value is to 1, the more accurate the trained model is indicated.
When the flow neural network prediction model NN1 is trained, the working rotation speed frequency f act and the motor input power P act are used as training input samples, the current working flow q est is used as a training output target result, and the calculation process and result of NN1 are as follows:
(1) The number of training sample data of NN1 is 396, and after training, the weight value between the hidden layer and the input layer is ：Wji＝[1.1764584879303381 1.7670120483614178;2.5488557147525275 -4.9817321534454093;-5.2740120297923809 4.4674727310219025;2.750307598034591 -4.6214585816646121;-0.72159347372304372-1.6371026285990855;3.0210992524645586 -1.4104891032380202;-5.6897614625943875-2.0703741590643987;0.3439243535980388 0.0758064616244499;-3.0612801844863098 4.5187664529366502;-2.864847162247627 2.2745472103221753];
(2) The weight value between the output layer and the hidden layer is ：Wkj＝[-1.2915940765276057 3.6816295973240463 1.9466345659934667-9.2684297529622199-2.1344293828585319 4.2133642780162823-0.15677411171366351-1.8779430546426432-5.3644864337478531 3.2635565312912949];
(3) The bias value of the hidden layer is ：θj＝[0.24808088916187218;-0.28906195440473975;-1.0079148663688584;-0.5329446756448406;-0.45358564993415079;1.5994469583024116;-4.3626337744891073;-0.63360296295822582;0.79149508031222515;-0.28598922506070112];
(4) The bias value of the output layer is: θ k = -0.67034513891577852;
the R value of the trained flow neural network prediction model is 0.99, and the model has higher precision.
When the pressure neural network prediction model NN2 is trained, the working rotation speed frequency f act and the current working flow q est are taken as training input samples, the current working pressure p est is taken as a training output target result, and the calculation process and result of NN2 are as follows:
(1) The number of training sample data of NN2 is 396, and after training, the weight value between the hidden layer and the input layer is ：Wml＝[-0.22871313296567705 1.3292062051807978;-0.3700712802676121 0.57977045519635595;-0.097993834617381759-0.73373538023873131;-0.20383891173436774 1.0823931365774999;0.7098774646841548 0.73298906904807992;-0.44268760908619414 1.2991720562000444;0.10672828230221042 0.13329536367118988;-0.64960039916400325 0.050664670728329655;0.5553717347997561 0.32999001239297071;0.042778559474070742 0.086690912468152109];
(2) The weight value between the output layer and the hidden layer is ：Wnm＝[-0.51468861150809553 -1.5358737006848464 1.1775587780814367 0.72881158549487901 0.21583463661180755 0.53783350533452112 0.067169122468895545-1.1770546996825009 0.47272050109166741 -0.043306044643445889];
(3) The bias value of the hidden layer is ：θm＝[0.38713314393923787;0.12161936447720154;0.75101491377046914;0.68338368247554981;-0.99989868650705149;0.3843513636540194;0.012729812612826331;1.0814380505427044;0.20256886806509583;0.029854748947845896];
(4) The bias value of the output layer is: θ n = -0.23963059359998504;
the R value of the trained pressure neural network prediction model is 0.99, and the model has higher precision.
The comparison between the front and back of the regulation after the regulation in this example is shown in table 1.
Table 1 centrifugal pump frequency conversion control front-back comparison based on double neural network model
After the regulation is completed, a new operation condition of the centrifugal pump 2 is obtained, the flow is 9.86m 3/h, the pressure is 502kPa, the rotating speed frequency of the motor 3 is 40.2Hz, the opening degree of the electric regulating valve 6 is 32%, the deviation between the regulated flow and the target value is 0.14m 3/h, and the deviation between the regulated pressure and the target value is 2kPa, which are all within the range of operation errors.
Embodiment two: see fig. 1, 3, 5:
this embodiment is substantially the same as the first embodiment described above, with the main differences: in step B1, the required operating flow value q set, the flow allowable error δ q, and the initial opening of the electric control valve for setting the centrifugal pump are deleted, step B5 and step B6 are deleted, and if the absolute value of the deviation between the current operating pressure p est and the required operating pressure p set is smaller than the pressure allowable error δ p in step B3, step B7 is entered.
In the embodiment, the opening control of the electric regulating valve 6 is deleted, the flow control is deleted, only the frequency control of the frequency converter 4 is reserved, the constant pressure control of the centrifugal pump 2 is realized through the rotation speed regulation of the motor 3, and the constant pressure control of a water supply system can be realized; the control strategy adopts closed-loop control, takes the input power P act and the working rotation speed frequency f act of the motor as the input parameters of the controller, outputs the current working pressure P est through a trained double-neural network prediction model, compares the pressure value with a set required working pressure value P set, outputs the adjusting rotation speed frequency f o to the frequency converter 4 through proportional pressure operation, and adjusts the working rotation speeds of the centrifugal pump 2 and the motor 3 through the frequency converter 4.
The above embodiments are only preferred embodiments of the present invention, and are not intended to limit the scope of the present invention in this way, therefore: all equivalent changes in structure, shape and principle of the invention should be covered in the scope of protection of the invention.
Claims (2)
1. A centrifugal pump variable frequency control method based on a double neural network model is characterized in that: the method comprises the following steps:
B1, setting a required working pressure value p set, a required working flow value q set, a pressure allowable error delta p, a flow allowable error delta q and an initial opening of an electric regulating valve of the centrifugal pump;
B2, measuring the input power P act and the working rotation speed frequency f act of the motor, and obtaining the current working pressure P est and the current working flow q est of the centrifugal pump by means of a double-neural network prediction model;
b3, based on the deviation between the current working pressure p est and the required working pressure value p set, the controller outputs the regulated rotating speed frequency f o, the regulated rotating speed frequency f o is used as the input frequency of the frequency converter, and the working rotating speeds of the motor and the centrifugal pump are controlled;
b4, judging whether the absolute value of the deviation between the current working pressure p est and the required working pressure p set is smaller than a pressure allowable error delta p, if so, entering a step B5; if not, returning to the step B2, and performing iteration control;
B5, based on the deviation between the current working flow q est and the required working flow q set, the controller outputs and adjusts the opening phi o of the electric regulating valve, takes the opening phi o of the electric regulating valve as an input value of an executing mechanism of the electric regulating valve, and controls the opening of the electric regulating valve;
B6, judging whether the absolute value of the deviation of the current working pressure p est and the required working pressure p set is smaller than a pressure allowable error delta p, judging whether the absolute value of the deviation of the current working flow q est and the required working flow q set is smaller than a flow allowable error delta q, and if the absolute value of the deviation of the current working pressure p est and the required working pressure p set are both satisfied, entering a step B7; if at least one item does not meet, returning to the step B2, and performing iterative control;
B7, finishing regulation;
The dual-neural network prediction model adopts a three-layer counter-propagation BP neural network and comprises a flow neural network prediction model NN1 and a pressure neural network prediction model NN2, wherein an input layer of the flow neural network prediction model NN1 comprises 2 neurons which are respectively a motor input power P act and a working rotating speed frequency f act, a hidden layer of the flow neural network prediction model NN1 comprises a plurality of neurons, and an output layer of the flow neural network prediction model NN1 comprises 1 neuron and is a current working flow q est; the input layer of the pressure neural network prediction model NN2 comprises 2 neurons, namely a working rotation speed frequency f act and a current working flow q est, the hidden layer of the pressure neural network prediction model NN2 comprises a plurality of neurons, the output layer of the pressure neural network prediction model NN2 comprises 1 neuron, and the current working pressure p est is obtained;
Acquiring a centrifugal pump hydraulic performance test dataset, wherein the centrifugal pump hydraulic performance test dataset comprises a current working pressure P est, a current working flow q est, a motor input power P act and a working rotating speed frequency f act, training a flow neural network prediction model NN1 and a pressure neural network prediction model NN2 respectively based on the centrifugal pump hydraulic performance test dataset, and distributing test data into a training sample dataset, a verification sample dataset and a test sample dataset according to the proportions of 80%, 10% and 10% by adopting a random distribution method;
The training algorithm adopts a Bayesian regularization algorithm, the iteration number of training is set to 5000, the learning rate is set to 0.1, the target error is set to 0.00001, and after training is completed, correlation analysis between target values and output values is carried out on a training result, a verification result and a test result respectively;
The hidden layer of the flow neural network prediction model NN1 comprises 10 neurons, and the hidden layer of the pressure neural network prediction model NN2 comprises 10 neurons;
The controller comprises a pressure PI controller and a flow PI controller, the pressure PI controller outputs a rotation speed frequency f o to control the working rotation speeds of the motor and the centrifugal pump; the flow PI controller outputs and adjusts the opening phi o of the electric regulating valve to control the opening of the electric regulating valve.
2. The centrifugal pump variable frequency control method based on the dual neural network model as claimed in claim 1, wherein the method is characterized by comprising the following steps: in step B2, the motor input power P act and the operating rotational speed frequency f act are read by a frequency converter or measured by a power sensor and a hall sensor connected to the motor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111340207.XA CN114046259B (en) | 2021-11-12 | 2021-11-12 | Centrifugal pump variable frequency control method based on double neural network model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111340207.XA CN114046259B (en) | 2021-11-12 | 2021-11-12 | Centrifugal pump variable frequency control method based on double neural network model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114046259A CN114046259A (en) | 2022-02-15 |
| CN114046259B true CN114046259B (en) | 2024-05-28 |
Family
ID=80208775
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111340207.XA Active CN114046259B (en) | 2021-11-12 | 2021-11-12 | Centrifugal pump variable frequency control method based on double neural network model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114046259B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114576554B (en) * | 2022-05-06 | 2022-08-02 | 浙江浙能航天氢能技术有限公司 | Hydrogen compression module control method and device based on neural network |
| CN115342063A (en) * | 2022-09-02 | 2022-11-15 | 浙江南元泵业有限公司 | Centrifugal pump energy consumption prediction method based on empirical mode decomposition and BP neural network |
| CN116378974B (en) * | 2023-05-31 | 2023-08-04 | 宜宾科全矿泉水有限公司 | Intelligent control system of water purifier |
| CN117540163A (en) * | 2023-11-09 | 2024-02-09 | 南栖仙策(南京)高新技术有限公司 | Pump performance curve generation method, model construction and training method and device |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104563217A (en) * | 2014-12-24 | 2015-04-29 | 卧龙电气集团股份有限公司 | Constant-pressure water supply controlling method without pressure sensor |
| CN105550447A (en) * | 2015-12-14 | 2016-05-04 | 渤海大学 | Dual neural network based modeling method for centrifugal water pump system |
| CN105839711A (en) * | 2016-03-24 | 2016-08-10 | 嘉兴极致传动科技有限公司 | Sensor-less variable frequency constant-pressure water supply control method |
| JP2017031883A (en) * | 2015-07-31 | 2017-02-09 | 株式会社日立産機システム | Pump device and operation control method for the same |
| CN108268060A (en) * | 2018-01-26 | 2018-07-10 | 温州大学 | A kind of flow control system and control method |
| CN111255756A (en) * | 2020-03-10 | 2020-06-09 | 常熟理工学院 | Variable speed pump control system in hydraulic system |
| CN111350651A (en) * | 2020-03-12 | 2020-06-30 | 利欧集团浙江泵业有限公司 | Constant-pressure control method of intelligent variable-frequency water pump |
| CN111859562A (en) * | 2020-07-10 | 2020-10-30 | 大连理工大学 | Pressure-independent intelligent regulating valve design method based on dynamic resistance characteristic |
-
2021
- 2021-11-12 CN CN202111340207.XA patent/CN114046259B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104563217A (en) * | 2014-12-24 | 2015-04-29 | 卧龙电气集团股份有限公司 | Constant-pressure water supply controlling method without pressure sensor |
| JP2017031883A (en) * | 2015-07-31 | 2017-02-09 | 株式会社日立産機システム | Pump device and operation control method for the same |
| CN105550447A (en) * | 2015-12-14 | 2016-05-04 | 渤海大学 | Dual neural network based modeling method for centrifugal water pump system |
| CN105839711A (en) * | 2016-03-24 | 2016-08-10 | 嘉兴极致传动科技有限公司 | Sensor-less variable frequency constant-pressure water supply control method |
| CN108268060A (en) * | 2018-01-26 | 2018-07-10 | 温州大学 | A kind of flow control system and control method |
| CN111255756A (en) * | 2020-03-10 | 2020-06-09 | 常熟理工学院 | Variable speed pump control system in hydraulic system |
| CN111350651A (en) * | 2020-03-12 | 2020-06-30 | 利欧集团浙江泵业有限公司 | Constant-pressure control method of intelligent variable-frequency water pump |
| CN111859562A (en) * | 2020-07-10 | 2020-10-30 | 大连理工大学 | Pressure-independent intelligent regulating valve design method based on dynamic resistance characteristic |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114046259A (en) | 2022-02-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114046259B (en) | Centrifugal pump variable frequency control method based on double neural network model | |
| CN111256294B (en) | Model prediction-based optimization control method for combined operation of water chilling unit | |
| CN103133319B (en) | A kind of condensing generator set Condensate Pump Frequency Conversion controlling method and system | |
| CN103670750B (en) | Power limit matching control system, method, device and engineering machinery | |
| JP2010506094A (en) | Wind turbine control system | |
| CN201292954Y (en) | Water pump and/or blower fan highly effective energy-conserving control system | |
| CN103711598B (en) | Regulation of hydraulic system equipment, method, Power Matching Control System and engineering machinery | |
| CN107870564B (en) | Anti-interference control method for fuel cell with decoupling performance | |
| CN108287571B (en) | Method for judging running interval of flow control system pump | |
| CN103953503A (en) | Yaw braking torque control device of wind generating set and method thereof | |
| CN108631362A (en) | A kind of wind power control method | |
| CN101994704B (en) | Chemical process intelligent pump and control method thereof | |
| CN114294159A (en) | Control method and device for variable-pitch double controllers of wind turbine generator | |
| CN115807729A (en) | Method and system for controlling power generation by using circulating water excess pressure | |
| CN108549207A (en) | A kind of method of Adaptive System of Water-Turbine Engine control parameter | |
| CN116241441A (en) | Automatic pump combining control method for steam feed pump of boiler water feed system | |
| CN107628691B (en) | Aeration sewage treatment system and multi-parameter fuzzy control method thereof | |
| Gevorkov et al. | Simulation study of mixed pressure and flow control systems for optimal operation of centrifugal pumping plants | |
| CN114383174B (en) | Unit control method and device and unit | |
| CN113836785B (en) | Municipal area intelligent water supply system and artificial intelligent control optimization method thereof | |
| CN1668829A (en) | Method and device for regulating the rotational speed of a turbine connected to an electric power supply mains by means of a generator | |
| CN111911907B (en) | Heater water level control quality evaluation method based on normalized error integral | |
| CN101701531B (en) | Overflow steam extraction adjusting device of industrial drive steam turbine and control method thereof | |
| CN114396826A (en) | Method for analyzing influence of accident state of water turbine for energy conservation of industrial circulating water on system | |
| CN113728162B (en) | Electronic device and method for optimizing use of motor-driven equipment in a control loop system |
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 |