Counter-propagation neural network centrifugal pump senseless prediction method based on improved sparrow search algorithm optimization
By improving the backpropagation neural network method optimized by the sparrow search algorithm, the problem of insufficient accuracy and local optimization of centrifugal pumps under flow and pressure sensor conditions is solved, and high-precision flow and head prediction is achieved, ensuring the operation of the equipment Safety and reliability.
Patent Information
- Application Number
- CN202410888271.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-04
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-07-04
AI Technical Summary
The existing centrifugal pump performance prediction methods have insufficient accuracy and local optimality problems under flow and pressure sensor conditions, and cannot achieve accurate flow and head prediction.
Using the backpropagation neural network method based on the optimization of the improved sparrow search algorithm, sample data is obtained through the hydraulic performance test of centrifugal pumps, mixed strategies are used to optimize the neural network model, and four-layer backpropagation neural network prediction model is established to carry out large-scale Training scale to achieve accurate prediction of flow and head.
The prediction error accumulation and local optimization problems of traditional methods are overcome, and high-precision flow and head prediction under flow and pressure sensor conditions are achieved, ensuring the safety and reliability of equipment operation.
Smart Images

Figure CN118939957A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of pump performance prediction methods, and specifically relates to a centrifugal pump performance neural network prediction method without a flow sensor, which is mainly used to quickly and accurately predict the working performance parameters of the centrifugal pump, accurately predict the flow rate and head of the pump without flow and pressure sensors, realize intelligent monitoring and diagnosis of the equipment operation status, and ensure the safety and reliability of the equipment operation. Background Art
[0002] For centrifugal pump systems, in order to ensure the safety and reliability of system operation, it is necessary to grasp the operating status of the equipment in real time and predict equipment operation failures. However, due to the limitations of physical space or special working environments, it is impossible to obtain information about the actual equipment operation by installing external sensors in the existing system, and external sensors will increase the complexity and cost of the system. At the same time, for some special working conditions, such as high temperature, high pressure, strong corrosion and other working media, it is impossible to monitor the operating status of the centrifugal pump by installing sensors. To solve this problem, some researchers have proposed a model-based prediction method, which combines the QP curve and the system pipeline characteristic curve to predict the flow and head of the centrifugal pump. However, this method is sensitive to the shape of the QP curve and the system pipeline characteristic curve, that is, the polynomial fitting equations of flow-power and flow-head, so there are certain defects in the prediction accuracy of flow and head. On this basis, the researchers developed a multi-level hybrid prediction model based on machine learning. This method aims at the problem of monitoring the performance parameters of the pump under the condition of no flow and pressure sensors. By establishing a multi-level JGPR model with speed and valve opening as input, the pump performance parameters can be predicted. However, since this method obtains the predicted value of the performance parameter by a step-by-step transmission method, it will lead to the accumulation of errors at each level during its transmission process. In view of the above problems, the present invention proposes a sensorless prediction method for centrifugal pumps based on a back propagation neural network optimized by an improved sparrow search algorithm. This method breaks away from the constraints of traditional polynomial fitting equations, uses a neural network model optimized by an intelligent algorithm, and conducts large-scale training on random sample data to obtain a centrifugal pump performance prediction model with higher accuracy and wider applicability, so as to achieve accurate prediction of centrifugal pump performance without flow and pressure sensors.
[0003] The existing performance prediction methods for centrifugal pumps without flow and pressure sensors have the following shortcomings: 1) For traditional model-based prediction methods, since the establishment of the model relies on QP curve, QH curve and system pipeline characteristic curve, the shape of these curves can greatly affect the prediction performance of the model, and it is impossible to effectively predict the flow and head; 2) The weights and thresholds of traditional neural networks are prone to fall into local optimality, resulting in inaccurate models generated and ineffective predictions. Summary of the invention
[0004] In order to overcome the shortcomings of the existing technology, the present invention provides a sensorless prediction method for centrifugal pumps based on a back propagation neural network optimized by an improved sparrow search algorithm. For a specific centrifugal pump system, relevant test data of the centrifugal pump is obtained in the laboratory with the help of a pump hydraulic performance test system, and the back propagation neural network is optimized by using an improved sparrow search algorithm with a hybrid strategy. The improved neural network model is used to perform large-scale training on the sample data. Through this prediction model, the flow and pressure parameters of the pump are accurately predicted without flow and pressure sensors, thereby realizing intelligent monitoring and diagnosis of the equipment operation status and ensuring the safety and reliability of the equipment operation.
[0005] The technical solution adopted by the present invention to solve its technical problem is:
[0006] A sensorless prediction method for centrifugal pumps based on a back propagation neural network optimized by an improved sparrow search algorithm comprises the following steps:
[0007] Step 1. Through the centrifugal pump hydraulic performance test, obtain the centrifugal pump flow Q, head H, motor shaft power P test sample data at different speeds n and valve openings V;
[0008] Step 2. Improve the sparrow search algorithm by hybrid strategy;
[0009] Step 3. Use the improved sparrow search algorithm to optimize the initial weights and thresholds of the back propagation neural network, and establish a back propagation neural network prediction model optimized by the improved sparrow search algorithm;
[0010] Step 4. Implant the trained prediction model into the centrifugal pump controller to achieve accurate prediction of centrifugal pump performance based on real-time measurement data.
[0011] Furthermore, in step 1, a centrifugal pump characteristic test is carried out under laboratory conditions, and the centrifugal pump flow Q, head H, and motor shaft power P at different speeds n and valve openings V are obtained with the help of a centrifugal pump hydraulic performance test system, the test data are sorted, and the centrifugal pump flow-head curve and flow-power curve at different speeds are plotted.
[0012] Furthermore, the process of step 2 is as follows:
[0013] First, the Logistic-Tent chaotic mapping is used to optimize the initial population generation rule to make the initial population distribution more uniform. The Logistic-Tent chaotic mapping is shown in formula (1); the initial population is initialized as shown in formula (2);
[0014]
[0015] x j =xlb +(x ub -x lb )×e j (2)
[0016] In the formula, α and γ are control factors, with values of 0.5 and 0.1 respectively, j is the dimension, mod is the modulus operation, and x lb is the lower bound of each population in each dimension, x ub is the upper bound of each population in each dimension, e j is the chaotic sequence generated by formula (1);
[0017] Secondly, the producer ratio is optimized using formula (3), and the adaptive control factor generated by formula (4) is used to improve the producer location distribution strategy, so that it can continue to explore near the local optimum and enhance the local optimal ability;
[0018]
[0019] In the formula, t is the current iteration number, iter max is the maximum number of iterations;
[0020] When the sparrow colony is in danger, the producer needs to move randomly. The Lévy flight strategy is introduced to improve the producer position update strategy. The Lévy flight strategy is shown in formula (5); the improved producer position update strategy is shown in formula (6);
[0021]
[0022] In the formula, β∈(1,3) is a random number, is the value of the jth dimension of the ith sparrow at the tth iteration, α∈(0,1] is a random number, R∈[0,1] is the alarm value, ST∈[0.5,1] is the safety threshold, μ is the step size scaling factor, is the dot product;
[0023] Finally, the Cauchy mutation disturbance is introduced to make the sparrow population trapped in the local optimum quickly jump out of the local optimum, as shown in formula (7);
[0024]
[0025] In the formula, x i is the current sparrow position, x best is the current global optimal position, x i ' is the new position after the Cauchy mutation disturbance, Cauchy(0,1) is a random variable that obeys the standard Cauchy distribution; is the dot product;
[0026] The population size of the hybrid strategy improved sparrow search algorithm (HSSA) is set to 30, the maximum number of iterations is set to 100, the safety threshold is set to 0.5, and the upper and lower boundaries are -3 and 3 respectively.
[0027] Furthermore, the process of step 3 is as follows:
[0028] Firstly, the motor speed, motor shaft power and valve opening are used as input parameters, and the sparrow search algorithm improved by hybrid strategy is combined to establish the centrifugal pump flow head neural network prediction model HSSA-BPNN. The neural network prediction model adopts a four-layer back propagation neural network method, which consists of an input layer, two hidden layers and an output layer. The input layer contains 3 neurons, namely motor speed, motor shaft power and valve opening, the first hidden layer contains 7 neurons, the second hidden layer contains 6 neurons, and the output layer contains 2 neurons, which are the flow and head outputs.
[0029] Secondly, the test measurement sample values (flow rate, head, motor speed, motor shaft power and valve opening) obtained in step 1 are used as the initial samples for neural network training to train the HSSA-BPNN prediction model. During training, the sample random allocation method is used to allocate the training samples into training sample data sets, verification sample data sets and test sample data sets in a ratio of 70%, 15% and 15%. The neural network training algorithm adopts the Levenberg-Marquardt algorithm, the learning rate is set to 0.01, the target error is set to 0.00001, and the maximum number of epochs is set to 160. After the training is completed, the correlation analysis between the target value and the output value is performed on the training results, verification results and test results, that is, R value analysis. The closer the R value is to 1, the more accurate the trained model is.
[0030] In step 4, the neural network prediction model selected by the centrifugal pump under different power characteristics is finally determined through step 3, and the trained neural network prediction model is implanted into the centrifugal pump controller, which is a standard programmable logic controller PLC. The electrical signals of the motor frequency, motor shaft power and valve opening in the frequency converter are collected and converted into real values of the motor frequency, motor shaft power and valve opening, and the motor speed is calculated using formula (8); the HSSA-BPNN prediction model is used to predict the flow value and head value under the current state, and to accurately predict the equipment performance and intelligently monitor and diagnose the equipment operation status;
[0031] n=60f (8) In the formula, n is the motor speed and f is the motor frequency.
[0032] The beneficial effects of the present invention are mainly manifested in: 1) with the help of the neural network model, through a large amount of sample data, a centrifugal pump performance prediction model is quickly established, and the initial weights and thresholds of the neural network are optimized using a sparrow search algorithm improved by a hybrid strategy to improve the performance of the neural network. This method does not need to establish a traditional centrifugal pump performance prediction mathematical model based on a polynomial fitting equation. At the same time, it overcomes the problems of large prediction errors and neural network parameters being trapped in local optimality and unable to jump out of the traditional mathematical model, and is significantly improved in prediction accuracy and applicability; 2) with the help of a controller, the flow and pressure of the pump are accurately predicted without flow and pressure sensors, realizing intelligent monitoring and diagnosis of the equipment operation status, and ensuring the safety and reliability of equipment operation. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Figure 1 Flow chart of the centrifugal pump non-sensing prediction method that jointly improves the sparrow search algorithm and the neural network algorithm.
[0034] Figure 2 Schematic diagram of the flow-head performance curve of a centrifugal pump at different speeds and frequencies.
[0035] Figure 3 Schematic diagram of the flow-power performance curve of a centrifugal pump at different speeds and frequencies.
[0036] Figure 4 Schematic diagram for comparing the sparrow population generated by the Logistic-Tent chaotic mapping method and the pseudo-random method.
[0037] Figure 5 Schematic diagram of the back propagation neural network optimized based on the improved sparrow search algorithm. DETAILED DESCRIPTION
[0038] The present invention will be further described below in conjunction with the accompanying drawings.
[0039] Reference Figure 1 to Figure 5 , A sensorless prediction method for centrifugal pumps based on back propagation neural network optimized by improved sparrow search algorithm, Figure 1 The flowchart of the centrifugal pump performance prediction method based on neural network is as follows: (1) Through the centrifugal pump hydraulic performance test, the centrifugal pump flow Q, head H, and motor shaft power P test sample data under different speeds n and valve openings V are obtained; (2) The sparrow search algorithm is improved by a hybrid strategy; (3) The initial weights and thresholds of the back propagation neural network are optimized using the improved sparrow search algorithm, and a back propagation neural network prediction model optimized by the improved sparrow search algorithm is established; (4) The trained prediction model is implanted into the centrifugal pump controller to accurately predict the flow and head of the centrifugal pump based on the real-time measured motor speed n, motor shaft power P, and valve opening V.
[0040] The sensorless prediction method for centrifugal pumps based on the back propagation neural network optimized by the improved sparrow search algorithm of this embodiment includes the following steps:
[0041] Step 1. Through the centrifugal pump hydraulic performance test, obtain the centrifugal pump flow Q, head H, motor shaft power P test sample data at different speeds n and valve openings V;
[0042] In this embodiment, a rated flow rate of Q n =6m 3 / h, rated head is H n =22m, rated speed is n=2900r / min, rated power is P n =0.75kW centrifugal pump as the test object, with the help of centrifugal pump hydraulic performance test system to obtain the centrifugal pump flow Q, head H, motor shaft power P measurement values at different speeds n and valve openings V, collate the test data, centrifugal pump performance data such as Figure 2 and Figure 3 As shown;
[0043] Step 2. Improve the sparrow search algorithm through a hybrid strategy. The process is as follows:
[0044] First, the Logistic-Tent chaotic mapping is used to optimize the initial population generation rule to make the initial population distribution more uniform. The Logistic-Tent chaotic mapping is shown in formula (1); the initial population is initialized as shown in formula (2); Figure 4 Comparison of sparrow populations generated by Logistic-Tent chaotic mapping method and pseudo-random method;
[0045]
[0046] x j =X tb +(x ub -x lb )×e j (2)
[0047] In the formula, j is the dimension, mod is the modulus operation, and x lb is the lower bound of each population in each dimension, x ub is the upper bound of each population in each dimension, e j is the chaotic sequence generated by formula (1);
[0048] Secondly, the producer ratio is optimized using formula (3), and the adaptive control factor generated by formula (4) is used to improve the producer location distribution strategy, so that it can continue to explore near the local optimum and enhance the local optimal ability;
[0049]
[0050] Where t is the current iteration number;
[0051] When the sparrow colony is in danger, the producer needs to move randomly. The Lévy flight strategy is introduced to improve the producer position update strategy. The Lévy flight strategy is shown in formula (5); the improved producer position update strategy is shown in formula (6);
[0052]
[0053] In the formula, β∈(1,3) is a random number, is the value of the jth dimension of the ith sparrow at the tth iteration, α∈(0,1] is a random number, R∈[0,1] is the alarm value, μ is the step size scaling factor, is the dot product;
[0054] Finally, the Cauchy mutation disturbance is introduced to make the sparrow population trapped in the local optimum quickly jump out of the local optimum, as shown in formula (7);
[0055]
[0056] In the formula, x i is the current sparrow position, x best is the current global optimal position, x i ' is the new position after the Cauchy mutation disturbance, Cauchy(0,1) is a random variable that obeys the standard Cauchy distribution; is the dot product;
[0057] Step 3. Use the improved sparrow search algorithm to optimize the initial weights and thresholds of the back propagation neural network, and establish a back propagation neural network prediction model optimized by the improved sparrow search algorithm. The process is as follows:
[0058] Firstly, the motor speed, motor shaft power and valve opening are used as input parameters, and the sparrow search algorithm improved by hybrid strategy is combined to establish the centrifugal pump flow head neural network prediction model (HSSA-BPNN). The neural network prediction model adopts a four-layer back propagation neural network method, which consists of an input layer, two hidden layers and an output layer. The input layer contains 3 neurons, namely motor speed, motor shaft power and valve opening, the first hidden layer contains 7 neurons, the second hidden layer contains 6 neurons, and the output layer contains 2 neurons, which are the flow and head outputs. The specific structure of the HSSA-BPNN prediction model is as follows: Figure 5 As shown;
[0059] Secondly, the test measurement sample values (flow rate, head, motor speed, motor shaft power and valve opening) obtained in step 1 are used as the initial samples for neural network training to train the HSSA-BPNN prediction model; during training, the sample random allocation method is used to allocate the training samples into training sample data sets, verification sample data sets and test sample data sets in a ratio of 70%, 15% and 15%, and the neural network training algorithm adopts the Levenberg-Marquardt algorithm, the learning rate is set to 0.01, the target error is set to 0.00001, and the maximum number of epochs is set to 160; after the training is completed, the correlation analysis between the target value and the output value is performed on the training results, verification results and test results, that is, R value analysis, and the closer the R value is to 1, the more accurate the trained model is;
[0060] The calculation process and results of the back propagation neural network prediction model HSSA-BPNN optimized by the improved sparrow search algorithm are as follows:
[0061] 1) The training sample data of HSSA-BPNN is 240, and the structure of a single sample data is an array of 3 inputs and 2 outputs; after training, the weight value between the first hidden layer and the input layer is:
[0062] W ji =[1.741166096 0.689774753 1.692645289;
[0063] 0.335237832 0.334847852 -0.406421296;
[0064] -1.520940796 1.060233026 0.359878596;
[0065] -1.808333881 3.422829066 0.293830819;
[0066] -1.635005196 0.585757087 1.970597506;
[0067] -0.651773841 -0.611348856 0.28831957;
[0068] -1.263497296 1.122956082 -1.431577099];
[0069] 2) The weight value between the second hidden layer and the first hidden layer is: W kj=[0.418787441-0.414098525 -0.569345584 -0.141670568 0.935483608 -1.79642941 -0.439618194; 0.255239077 -1.928998206 0.084891585 0.299867508 -0.06146623 1.864425455-1.105248012; -0.639530178 -0.52963304 -1.193898612 2.063914289 -1.7483678922.0686502281.907800076; -0.629882699 1.411750601 0.172836529 0.401125578
[0070] 0.66911226-0.796317714 -2.887723481; -0.559033952 -0.376257183 -2.9020248772.047420934 -2.324464603 5.575811625 0.804380083; 0.0365759291.184496636-0.778223325 0.197075478 -0.104708648 -2.565439718 -0.763185825];
[0071] 3) The weight value between the output layer and the second hidden layer is:
[0072] W lk =[0.265215932 -0.268822059 3.458677935 0.157833232 -2.320218120.59875929;
[0073] 0.174076823 -0.453687127 -1.957035347 0.023024239 0.9446206720.642233047];
[0074] 4) The bias value of the first hidden layer is:
[0075] θ j =[-1.292357635;
[0076] -0.986754666;
[0077] -0.029579731;
[0078] -1.766221063;
[0079] -0.659089594;
[0080] -1.530234027; 1.018581205];
[0082] 5) The bias value of the second hidden layer is:
[0083] θ k =[-1.540808196; 1.064614859;
[0085] -1.86744518; 2.644866503; 1.393608489;
[0088] -2.11527632];
[0089] 6) The deviation value of the output layer is:
[0090] θ l =[1.921565735;
[0091] -0.832669931];
[0092] The R values of the flow rate and head of the trained model are 0.99901 and 0.9997, respectively, indicating that the model has high accuracy;
[0093] Step 4. Implant the trained prediction model into the centrifugal pump controller to achieve accurate prediction of centrifugal pump performance based on real-time measurement data;
[0094] Through step 3, the neural network prediction model selected for the centrifugal pump under different power characteristics is finally determined. The trained neural network prediction model is implanted into the centrifugal pump controller. The controller is a standard programmable logic controller (PLC). The electrical signals of the motor frequency, motor shaft power and valve opening in the inverter are collected and converted into real values of the motor frequency, motor shaft power and valve opening. The motor speed is calculated using formula (8). The HSSA-BPNN prediction model is used to predict the flow value and head value under the current state, and to achieve accurate prediction of equipment performance and intelligent monitoring and diagnosis of equipment operation status.
[0095] n=60f (8) In the formula, n is the motor speed and f is the motor frequency.
[0096] Table 1 shows the real-time measurement data of the centrifugal pump
[0097]
[0098] Table 1
[0099] Table 2 is a comparative analysis of the predicted performance and measured performance of the centrifugal pump
[0100]
[0101] Table 2
[0102] Finally, the real-time flow, head and efficiency operating parameter values were obtained through the neural network prediction model. In the absence of flow sensors, the performance prediction of the centrifugal pump and the real-time monitoring of the equipment operating status were realized to ensure the safety and reliability of the equipment operation.
[0103] The contents described in the embodiments of this specification are merely enumerations of implementation forms of the inventive concept and are for illustrative purposes only. The protection scope of the present invention should not be considered to be limited to the specific forms described in this embodiment, and the protection scope of the present invention also extends to equivalent technical means that can be thought of by ordinary technicians in this field based on the inventive concept.
Claims
1. A sensorless prediction method for centrifugal pumps based on a back propagation neural network optimized by an improved sparrow search algorithm, characterized in that: The method comprises the following steps: Step 1. Through the centrifugal pump hydraulic performance test, obtain the centrifugal pump flow Q, head H, motor shaft power P test sample data at different speeds n and valve openings V; Step 2. Improve the sparrow search algorithm by hybrid strategy; Step 3. Use the improved sparrow search algorithm to optimize the initial weights and thresholds of the back propagation neural network, and establish a back propagation neural network prediction model optimized by the improved sparrow search algorithm; Step 4. Implant the trained prediction model into the centrifugal pump controller to achieve accurate prediction of centrifugal pump performance based on real-time measurement data.
2. The sensorless prediction method for centrifugal pumps based on the back propagation neural network optimized by the improved sparrow search algorithm according to claim 1 is characterized in that: In the step 1, a centrifugal pump characteristic test is carried out under laboratory conditions, and the centrifugal pump flow Q, head H, and motor shaft power P at different speeds n and valve openings V are obtained with the help of a centrifugal pump hydraulic performance test system. The test data are sorted out and the centrifugal pump flow-head curve and flow-power curve at different speeds are plotted.
3. The sensorless prediction method for centrifugal pumps based on the back propagation neural network optimized by the improved sparrow search algorithm according to claim 1 or 2, characterized in that: The process of step 2 is as follows: First, the Logistic-Tent chaotic mapping is used to optimize the initial population generation rule to make the initial population distribution more uniform. The Logistic-Tent chaotic mapping is shown in formula (1); the initial population is initialized as shown in formula (2); x j =x lb +(x ub -x lb )×e j (2) In the formula, α and γ are control factors, with values of 0.5 and 0.1 respectively, j is the dimension, mod is the modulus operation, and x lb is the lower bound of each population in each dimension, x ub is the upper bound of each population in each dimension, e j is the chaotic sequence generated by formula (1); Secondly, the producer ratio is optimized using formula (3), and the adaptive control factor generated by formula (4) is used to improve the producer location distribution strategy, so that it can continue to explore near the local optimum and enhance the local optimal ability; In the formula, t is the current iteration number, iter max is the maximum number of iterations; When the sparrow colony is in danger, the producer needs to move randomly. The Lévy flight strategy is introduced to improve the producer position update strategy. The Lévy flight strategy is shown in formula (5); the improved producer position update strategy is shown in formula (6); In the formula, β∈(1,3) is a random number, is the value of the jth dimension of the ith sparrow at the tth iteration, α∈(0,1] is a random number, R∈[0,1] is the alarm value, ST∈[0.5,1] is the safety threshold, μ is the step size scaling factor, is the dot product; Finally, the Cauchy mutation disturbance is introduced to make the sparrow population trapped in the local optimum quickly jump out of the local optimum, as shown in formula (7); In the formula, x i is the current sparrow position, x best is the current global optimal position, x i ’ is the new position after the Cauchy mutation perturbation, Cauchy(0,1) is a random variable that obeys the standard Cauchy distribution; is the dot product; The population size of the hybrid strategy improved sparrow search algorithm (HSSA) is set to 30, the maximum number of iterations is set to 100, the safety threshold is set to 0.5, and the upper and lower boundaries are -3 and 3 respectively.
4. The sensorless prediction method for centrifugal pumps based on the back propagation neural network optimized by the improved sparrow search algorithm according to claim 1 or 2, characterized in that: The process of step 3 is as follows: Firstly, the motor speed, motor shaft power and valve opening are used as input parameters, and the sparrow search algorithm improved by hybrid strategy is combined to establish the centrifugal pump flow head neural network prediction model HSSA-BPNN. The neural network prediction model adopts a four-layer back propagation neural network method, which consists of an input layer, two hidden layers and an output layer. The input layer contains 3 neurons, namely motor speed, motor shaft power and valve opening, the first hidden layer contains 7 neurons, the second hidden layer contains 6 neurons, and the output layer contains 2 neurons, which are the flow and head outputs. Secondly, the experimental measurement sample values (flow rate, head, motor speed, motor shaft power and valve opening) obtained in step 1 are used as the initial samples for neural network training to train the HSSA-BPNN prediction model; During training, the random sample allocation method was used to allocate the training samples into training sample data sets, validation sample data sets, and test sample data sets in a ratio of 70%, 15%, and 15%. The neural network training algorithm used the Levenberg-Marquardt algorithm, the learning rate was set to 0.01, the target error was set to 0.00001, and the maximum number of epochs was set to 160; After the training is completed, the correlation analysis between the target value and the output value, i.e., R value analysis, is performed on the training results, verification results, and test results. The closer the R value is to 1, the more accurate the trained model is.
5. The sensorless prediction method for centrifugal pumps based on the back propagation neural network optimized by the improved sparrow search algorithm according to claim 1 or 2, characterized in that: In step 4, the neural network prediction model selected by the centrifugal pump under different power characteristics is finally determined through step 3, and the trained neural network prediction model is implanted into the centrifugal pump controller, which is a standard programmable logic controller PLC. The electrical signals of the motor frequency, motor shaft power and valve opening in the frequency converter are collected and converted into real values of the motor frequency, motor shaft power and valve opening, and the motor speed is calculated using formula (8); the HSSA-BPNN prediction model is used to predict the flow value and head value under the current state, and to accurately predict the equipment performance and intelligently monitor and diagnose the equipment operation status; n=60f (8) In the formula, n is the motor speed and f is the motor frequency.
Citation Information
Patent Citations
Power electronic circuit fault diagnosis method for optimizing ELM based on improved sparrow search algorithm
CN113269252A
Photovoltaic base station short-term photovoltaic power prediction method based on improved sparrow algorithm optimization
CN113762603A
Centrifugal pump performance neural network prediction method without flow sensing
CN114109859A
Direct current (DC) / DC converter fault diagnosis method and system based on improved sparrow search algorithm
US20230394316A1
