Intelligent pump station characteristic correlation fitting optimization method, equipment and medium

Abstract

The invention discloses an intelligent pump station feature correlation fitting optimization method and device and a medium, and relates to the field of data processing.The method comprises the steps that input features of a target water pump are input into a plurality of machine learning algorithm models to obtain a plurality of output features; the output characteristics comprise the blade angle and the operation efficiency of the target water pump; the input characteristics comprise lift and flow; the plurality of machine learning algorithm models comprise a preset support vector regression model, a preset BP neural network model and a preset random forest regression model; determining a final output feature according to the plurality of output features based on a preset decision rule and the model prediction error corresponding to each output feature; the correlation between the final output feature and the input feature is the optimized pump station feature correlation. According to the method, the operation efficiency of the model decision water pump and the adaptive capacity and prediction precision of the blade angle can be improved.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a method, device and medium for optimizing the feature correlation fitting of an intelligent pumping station. Background Art

[0002] As a key equipment for fluid transportation, water pumps play an indispensable role in various industrial, agricultural and civil fields. The operating efficiency of water pumps not only directly affects the energy consumption, service life and reliability of the equipment, but also has a vital impact on the economy and environmental friendliness of the entire system. In the actual operation of water pumps, it is necessary to obtain the corresponding blade angle parameters and adjust them according to the correlation between information such as flow rate and information such as operating efficiency to optimize the operating efficiency of the water pump and minimize the total power consumption.

[0003] There are currently two methods for determining the correlation between information such as flow rate and operating efficiency of a water pump under different working conditions: 1) Manual determination: Manually reading and determining parameters through the comprehensive characteristic curve of the water pump, but the parameter reading range of this manual processing method is small, the reading accuracy is low, and the real-time adjustment capability is poor. 2) Neural network model: The nonlinear relationship between input and output is obtained through a single neural network model, but this method has the problem of error uncertainty. New samples need to be continuously input to continuously approximate the relationship between input and output, and the structure and parameters of the network largely determine the accuracy of the network, that is, new samples may be continuously added, and the original structure and parameters cannot be trained into a model with smaller errors. In short, using only a single neural network model has the problems of poor robustness, weak self-correction ability, insufficient self-adjustment ability of the model when new samples are added, and weak intelligent ability. Summary of the invention

[0004] The purpose of this application is to provide a method, device and medium for optimizing the fitting of characteristic correlation of intelligent pumping stations, which can improve the operating efficiency of model-determined water pumps and the adaptability and prediction accuracy of blade angles.

[0005] To achieve the above objectives, this application provides the following solutions: In a first aspect, the present application provides a method for optimizing the correlation fitting of characteristics of an intelligent pumping station, comprising: Inputting the input features of the target water pump into a plurality of machine learning algorithm models respectively to obtain a plurality of output features; each of the output features includes the blade angle and the operating efficiency of the target water pump; the input features include the head and the flow rate; the plurality of machine learning algorithm models include a preset support vector regression model, a preset BP neural network model and a preset random forest regression model; Obtaining the model prediction error corresponding to each of the output features; Based on preset decision rules and the model prediction error corresponding to each of the output features, a final output feature is determined according to multiple output features; the correlation between the final output feature and the input feature is the optimized pump station feature correlation.

[0006] In a second aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a method for optimizing the fitting of characteristic correlations of intelligent pumping stations.

[0007] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for optimizing the fitting of characteristic correlations of intelligent pumping stations.

[0008] According to the specific embodiments provided by the present application, the present application has the following technical effects: the present application provides a method, device and medium for optimizing the correlation fitting of the characteristics of an intelligent pumping station, which combines three models, namely a preset support vector regression model, a preset BP neural network model and a preset random forest regression model, to achieve the nonlinear relationship fitting of the head and flow of the water pump, the blade angle and the operating efficiency, and determine the final output characteristics based on the model prediction errors and preset decision rules of the above three models, to achieve correction of the decision results, so that the final output characteristics are closer to the actual value, thereby improving the adaptive ability and prediction accuracy of the model decision. In short, in the present application, the generalization ability and accuracy of the model are improved through multiple model predictions and decision corrections, and it can be ensured that when new data is added to the sample, the intelligent prediction model has better adaptive, self-decision-making, self-learning, self-perception and adaptive capabilities, and realizes the self-adaptation and intelligent adjustment of the water pump unit, ensuring that it can operate efficiently, safely and stably under different working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0010] Figure 1 This is an application environment diagram of the intelligent pump station feature correlation fitting optimization method in one embodiment of the present application.

[0011] Figure 2 It is a flow chart of the intelligent pump station characteristic correlation fitting optimization method in one embodiment of the present application.

[0012] Figure 3Schematic diagram of the overall network framework of the intelligent pump station feature correlation fitting optimization method in one embodiment of the present application.

[0013] Figure 4 A schematic diagram of the structure of a preset decision model provided in one embodiment of the present application.

[0014] Figure 5 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0015] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0016] This application adopts a combination of multiple fitting regression techniques and uses a decision model to further correct the output of the model, thereby improving the accuracy of model parameter reading.

[0017] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0018] The intelligent pump station feature correlation fitting optimization method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, or it can be integrated on the server 104, or it can be placed on the cloud or other servers. The terminal 102 can send the input characteristics of the target water pump to the server 104. After the server 104 receives it, the input characteristics of the target water pump are respectively input into multiple machine learning algorithm models, and then the final output characteristics are determined according to the model prediction error corresponding to each output characteristic and the preset decision rules. The server 104 can feedback the final output characteristics to the terminal 102. In addition, in some embodiments, the intelligent pump station feature correlation fitting optimization method can also be implemented separately by the server 104 or the terminal 102.

[0019] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart vehicle-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or may be a cloud server.

[0020] In an exemplary embodiment, Figure 2 As shown, a method for optimizing the correlation fitting of intelligent pump station characteristics is provided. The method is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or can be executed by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used for explanation, and the steps include the following steps 201 to 203.

[0021] Step 201, input the input features of the target water pump into multiple machine learning algorithm models respectively to obtain multiple output features; each of the output features includes the blade angle and operating efficiency of the target water pump; the input features include head and flow; the multiple machine learning algorithm models include a preset support vector regression model, a preset BP neural network model and a preset random forest regression model.

[0022] The pump head refers to the energy obtained by a unit weight of fluid passing through the pump, which can also be understood as the work done by the pump on a unit weight of liquid, that is, the height to which the pump can lift water. To facilitate the expression of the subsequent model, the pump head and flow rate are expressed as and , the blade angle and operating efficiency of the water pump are expressed as and .

[0023] Step 202: Obtain the model prediction error corresponding to each of the output features.

[0024] like Figure 3 As shown, corresponding to step 201 and step 202, the overall network framework of the present application includes three branches, namely, a preset support vector regression model, a preset BP neural network model and a preset random forest regression model, and each model will output corresponding output features and model prediction errors. In practical applications, other models can be selected as needed, or the number of machine learning algorithm models can be increased, such as using four models, five models, etc., which can be set by relevant technical personnel.

[0025] The preset support vector regression model used in this application is a support vector regression (SVR), which is an extension of the support vector machine and can fit the regression model function that meets the requirements. This application uses the support vector regression model to map the samples to the high-dimensional space using the Gaussian kernel function to find the optimal hyperplane. This application needs to establish two models, the first one is the head and flow input, and the blade angle is the output; the second one is the head and flow input, and the operating efficiency is the output. Each pair of input features is represented as a vector of length 2 , including known input features (or historical input features) and The output features are , To facilitate the subsequent formula expression, the output features are expressed as y 1,2 The regression model is represented as: .

[0026] The optimization objective is expressed as: .

[0027] Where N is the number of historical data of water pumps in the training set, , , , is the Lagrange multiplier, is a Gaussian kernel function used to calculate the similarity between points; is the regularization parameter, which controls the degree of error penalty; For the parameters of the insensitive loss function, the allowable error range is defined; and is a slack variable used to process sample points that are beyond the error range; is the bias term. The Gaussian kernel function is expressed as: .

[0028] in, Enter data for the pump history dataset, is the point that needs to be predicted, represents the Euclidean distance between the two. Represents the bandwidth of the Gaussian distribution.

[0029] In a specific application, based on the above principle, the construction process of the preset support vector regression model includes the following steps (11)-(16).

[0030] (11) Obtaining a water pump historical data set; each water pump historical data in the water pump historical data set includes head, flow rate, blade angle and operating efficiency. Specifically, firstly, collecting and sorting the existing initial historical data, and performing pre-processing operations such as removing outliers, removing duplicates, standardizing, and regularizing the initial historical data, so as to obtain a water pump historical data set to be used.

[0031] (12) The water pump historical data set is divided into a training set and a test set, which are used to train model parameters and verify model parameters respectively. In practical applications, the ratio of 7:3 can be used, or it can be adjusted appropriately as needed.

[0032] (13) Construct the first support vector regression model and the second support vector regression model.

[0033] (14) Inputting the training set into the first support vector regression model, taking the lift and the flow rate as input and the blade angle as output, obtaining the optimal hyperplane by minimizing the objective function, so as to obtain a first initial support vector regression model.

[0034] (15) Input the training set into the second support vector regression model, take the lift and flow rate as input, take the operation efficiency as output, and obtain the optimal hyperplane by minimizing the objective function to obtain the second initial support vector regression model. In the training process of the above two support vector regression models, set the parameters , , , and the remaining unknown parameters are obtained through training.

[0035] (16) Using the test set to test and optimize the first initial support vector regression model and the second initial support vector regression model, respectively, to obtain two final support vector regression models that meet the first preset requirement; the two final support vector regression models constitute the preset support vector regression model.

[0036] In the actual test process, the data in the test set is input into the trained support vector regression model, and the error between the output value and the true value is evaluated. If the error is large, the parameters are adjusted. , and Until the requirements are met; after the training is completed, the head and flow that need to be predicted can be input into the corresponding model to obtain the predicted value.

[0037] The support vector regression algorithm has a solid theoretical foundation, is good at processing small sample data and high-dimensional data, has nonlinear mapping capabilities, is tolerant to errors and has strong robustness, and has flexible kernel function selection. By training the support vector regression model, a nonlinear relationship between head, flow rate, blade angle, and operating efficiency can be established, and the predicted value given by the support vector machine can be obtained as one of the reference output values.

[0038] The preset BP neural network (Back Propagation Neural Network) model used in this application is a multi-layer feedforward neural network model. Its core idea is the gradient descent method. It generates prediction values ​​through forward propagation, and uses back propagation to adjust the network weights according to the prediction error to minimize the loss function, thereby realizing the learning of the input-output mapping relationship. The BP neural network has strong self-learning and adaptive capabilities. It builds a neural network structure and appropriately adjusts the parameters to train the network structure, builds the relationship between input and output, and thus realizes the prediction function. Among them, the network prediction model can be expressed as: .

[0039] in, and is the weight matrix, b1 and b2 are the bias matrices, and is the activation function, Refers to, which can be expressed as: The number of neurons and layers in the BP neural network needs to be determined through debugging. After each training, the weight matrix and bias matrix are updated through the loss function (a non-negative real-valued function used to measure the degree of fit of the algorithm to the data, that is, the degree of inconsistency between the predicted value and the true value of the evaluation model) and the gradient descent method. The loss function is: .

[0040] in, f loss is the value of the loss function, w is the number of historical data of water pumps involved in the training of BP neural network, is the predicted value of BP neural network training, is the corresponding actual value. In order to minimize the loss value, update the weights in the opposite direction of the gradient and bias b, expressed as: .

[0041] in, is the learning rate, which needs to be determined based on empirical testing; weight Refers to the weight matrix and , bias b refers to the bias matrices b1 and b2.

[0042] In a specific application, based on the above principle, the construction process of the preset BP neural network model includes steps (21)-(25).

[0043] (21) Obtaining a water pump historical data set; each water pump historical data in the water pump historical data set includes head, flow, blade angle and operating efficiency; specifically, firstly collecting and organizing the existing initial historical data, normalizing the input data in the initial historical data, namely head and flow, and normalizing the existing output data, namely blade angle and operating efficiency, so as to obtain a water pump historical data set to be used. That is, corresponding to the following step (24), inputting the training set into the BP network includes: normalizing each water pump historical data in the training set, and then inputting it into the BP network.

[0044] (22) Divide the water pump historical data set into a training set and a test set. In practical applications, the ratio of the ratio can be 7:3, or it can be adjusted appropriately as needed.

[0045] (23) Build a BP network; specifically, determine the number of neurons and layers, and set the learning rate and number of training times.

[0046] (24) Inputting the training set into the BP network, taking the lift and the flow rate as input, taking the blade angle and the operating efficiency as output, and performing network training according to a preset learning rate and a preset number of training times, so as to obtain an initial BP neural network model. Specifically, adjusting parameters so that the loss of the BP network training meets the requirements. In addition, when adding new samples, generally, it is not necessary to modify the relevant parameters. When the loss is too large, it may be necessary to adjust the network structure and relevant parameters.

[0047] (25) Using the test set to test and optimize the initial BP neural network model to obtain a final BP neural network model that meets the second preset requirement; the final BP neural network model is the preset BP neural network model.

[0048] When predicting in actual applications, it is necessary to standardize the input features according to the data processing method during training. The neural network can quickly give outputs based on the trained weights and biases. The BP neural network has the ability to self-learn and can automatically extract features from the data for regression through training; the BP neural network can simulate complex nonlinear relationships through the connection and activation functions of multiple layers of neurons, so its accuracy has an extremely high upper limit; the BP network uses parallel processing to make calculations fast and efficient, which enables the neural network to maintain a high computing speed when processing large-scale data sets. By establishing a BP network model of head, flow, blade angle, and operating efficiency, the highest accuracy sample output value can be given in theory through training and parameter adjustment, so it can be used to improve output accuracy in the overall model.

[0049] The preset Random Forest Regression (RF) regression model used in this application is an ensemble learning method that combines the idea of ​​decision trees and the advantages of ensemble learning. It constructs multiple decision trees and averages the prediction results of each tree to perform regression prediction. The random forest regression algorithm has the advantages of strong robustness and generalization ability. In the sorted data (i.e., the historical data set of the water pump), a sample subset is randomly sampled: .

[0050] Each sample in the sample subset is represented as: . A decision tree is constructed for each sample subset. During the construction, some features are randomly selected as the final samples, and the samples are divided into two parts: .

[0051] Among them, j is the index, representing the jth sample in the sample subset, Represents the split point. Since the input is a two-dimensional point, the split point is an implicit point. The above expression can be interpreted as traversing all situations where the sample is divided into two parts, judging the objective function based on all situations, and finding a division method that satisfies the objective function. The objective function is expressed as: .

[0052] in, , is the average value of the output value of the partitioned area. Calculate all possible partitions until the objective function is minimized and the optimal split point is determined. Repeat the above steps for the two partitioned parts and further divide until the number of samples is less than the threshold. The minimum sample is the leaf node. After constructing the decision tree for each sample, we get A binary decision tree: .

[0053] When predicting, for the input features that need to be predicted , starting from each decision tree root node, determine the correlation with each point, determine the strongest correlation by minimizing the Euclidean distance, and then continue to traverse the nodes until the leaf node, then the average output value of the leaf node is the predicted value.

[0054] In a specific application, based on the above principles, the construction process of the preset random forest regression model includes the following steps (31)-(33).

[0055] (31) Obtaining a water pump historical data set; each water pump historical data in the water pump historical data set includes head, flow rate, blade angle and operating efficiency.

[0056] (32) Extracting a plurality of water pump historical data as samples with replacement from the water pump historical data set to obtain a sample subset. The amount of data contained in each sample in the sample subset may be different.

[0057] (33) For any sample in the sample subset, a binary decision tree is constructed using the CART (Classification and Regression Trees) algorithm; the binary decision trees corresponding to multiple samples constitute a preset random forest regression model. For newly added samples, there is no need to adjust parameters when re-establishing the random forest, which is very convenient.

[0058] In the actual prediction process, the required prediction value is input, and the final prediction value can be obtained by taking the average of the prediction values ​​given by each decision tree. Random forest regression can improve the stability and generalization ability of the model by combining the prediction results of multiple decision trees through ensemble learning. Even if there are noise or outliers in the training data, random forest can reduce these effects through majority voting or average prediction, thereby improving the accuracy of the prediction. In the entire model, random forest can give a stable decision solution. Even when the sample changes greatly and causes other models to be unstable, the model can still run normally and calculate a relatively moderate prediction value. By establishing a nonlinear relationship between the features of multiple decision trees between head, flow, blade angle, and operating efficiency, the output calculated by the random forest model can be given according to the input value that needs to be predicted.

[0059] In summary, the preset support vector regression model, the preset BP neural network model and the preset random forest regression model are expressed as: .

[0060] in, To preset the support vector regression model, To preset the random forest regression model, It is the preset BP network model; To refer to or , To preset the number of binary decision trees in the random forest regression model, is the i-th binary decision tree in the preset random forest regression model, are the input features that need to be predicted, namely head and flow rate.

[0061] like Figure 3 As shown, a support vector regression model, a BP neural network model and a random forest regression model are preset, and these three models correspond to three outputs and model prediction errors. After adding new samples, the prediction errors of each model are constantly changing. Therefore, the decision-making idea is to first determine whether the errors of the first three models are less than the threshold. If they are satisfied, select from the prediction results of the first three models, otherwise enter the decision model (DM) to further output the final result. The decision model is used to decide the output results of multiple models to make the results closer to the actual value, as shown in the following step 203.

[0062] Step 203, based on the preset decision rules, the model prediction error corresponding to each of the output features, and the multiple output features, the final output feature is determined; the correlation between the final output feature and the input feature is the optimized pump station feature correlation.

[0063] In an exemplary embodiment, after the three models are established, the advantages and disadvantages of each model are unknown during training, that is, it is unclear which model works best after adding new samples, and a preset decision model is proposed for more accurate prediction. Step 203 includes the following steps (31)-(33). Through the processing of these three steps, a decision model is established. Based on the predictions of the three algorithms that have been completed above, a quadruple prediction output can be performed, avoiding the problem that a single model cannot give a correct result when the error is large, thereby improving the model robustness, adaptability and decision accuracy.

[0064] (31) Determine whether the model prediction error corresponding to each of the output features is less than a preset error threshold (like ). Wherein, the model prediction error corresponding to each of the output features is the mean square error.

[0065] (32) If at least one model prediction error is less than a preset error threshold, then one or more output features whose model prediction error is less than the preset error threshold are averaged to obtain the final output feature. Specifically, when only one model prediction error is less than the preset error threshold, the following formula is used to determine the final output feature: .

[0066] in, Y pred is the final output feature, loss1 is the model prediction error corresponding to the preset support vector regression model, loss2 is the model prediction error of the preset random forest regression model, and loss3 is the model prediction error corresponding to the preset BP network model.

[0067] When the prediction error of one or more models is less than the preset error threshold, the following formula is used to determine the final output features: .

[0068] in, The number of output features for which the model prediction error is less than the preset error threshold.

[0069] (33) If the prediction errors of all models are not less than the preset error threshold, all the output features are input into the preset decision model to obtain the final output features; the preset decision model is a trained BP neural network. That is, the underlying principle of the decision model is the same as that of the BP neural network. Its purpose is to use the decision model to correct the results when the errors of the results given by the previous three models are large, so that the results are closer to the true value. If new samples are added, the above three models need to be retrained and different decision plans are given according to the errors of each model in each training.

[0070] In a specific application, such as Figure 4 As shown in the figure, for the three machine learning algorithm models that have been built and trained, the corresponding blade angle and operating efficiency values ​​are obtained through the head and flow rate, and the corresponding output value Y is obtained. SVR , Y RF , Y BP , and it is known that the errors of the three model training are , , , if at least one of the three model prediction errors is less than , take less than The average of the model outputs is used as output ; If the prediction errors of the three models are all greater than , the outputs of the three models are used as the input of the preset decision model, and the final result is calculated by the decision model as the output .

[0071] In addition, the training samples of the preset decision model come from the training results of the previous three models, and its model prediction error is less than Then it is used as training data, otherwise the group of data is discarded. The input and output of each group are expressed as: .

[0072] After obtaining the sample set to be used through the above screening, a BP neural network is established, including the input layer, the first fully connected layer, the second fully connected layer, the activation layer and the output layer, each fully connected layer contains 4 neurons, and the activation function selects the Sigmoid function. The final preset decision model obtained through training is expressed as: .

[0073] In summary, this application uses three machine learning algorithms to establish a nonlinear relationship model of head, flow rate, blade angle, and operating efficiency to avoid the problems of manually reading characteristic curves; this application proposes a decision model based on BP neural network. When the fitting errors of the three models (i.e., model prediction errors) are large, corrections are made through the decision model to make the output closer to the actual value, and to improve the model's adaptability and prediction accuracy. This application improves the generalization ability and accuracy of the model through the above-mentioned multi-model decision fitting method, and ensures that when new data is added to the sample, the model has better adaptability, self-decision-making, self-learning, self-perception, and self-adaptation capabilities, so that the pump station unit can be intelligently adjusted by itself, ensuring efficient, safe, and stable operation under different working conditions.

[0074] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 5 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it realizes the intelligent pump station feature correlation fitting optimization method.

[0075] Those skilled in the art will understand that Figure 5The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a method for optimizing the correlation fitting of characteristics of an intelligent pumping station.

[0076] In an exemplary embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, a method for optimizing the fitting of characteristics correlation of an intelligent pumping station is implemented.

[0077] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0078] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0079] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0080] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0081] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0082] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for optimizing the correlation fitting of characteristics of an intelligent pumping station, characterized in that: The method comprises: Inputting the input features of the target water pump into a plurality of machine learning algorithm models respectively to obtain a plurality of output features; each of the output features includes the blade angle and the operating efficiency of the target water pump; the input features include the head and the flow rate; the plurality of machine learning algorithm models include a preset support vector regression model, a preset BP neural network model and a preset random forest regression model; Obtaining the model prediction error corresponding to each of the output features; Based on preset decision rules and the model prediction error corresponding to each of the output features, a final output feature is determined according to multiple output features; the correlation between the final output feature and the input feature is the optimized pump station feature correlation.

2. The intelligent pump station characteristic correlation fitting optimization method according to claim 1 is characterized in that: Based on the preset decision rule and the model prediction error corresponding to each of the output features, the final output feature is determined according to the multiple output features, including: Determine whether the model prediction error corresponding to each of the output features is less than a preset error threshold; If at least one model prediction error is less than a preset error threshold, then average calculation is performed based on one or more output features whose model prediction error is less than the preset error threshold to obtain the final output feature; If the prediction errors of all models are not less than the preset error threshold, all the output features are input into the preset decision model to obtain the final output features; the preset decision model is a trained BP neural network.

3. The intelligent pump station characteristic correlation fitting optimization method according to claim 1 is characterized in that: The construction process of the preset support vector regression model includes: Acquire a water pump historical data set; each water pump historical data in the water pump historical data set includes head, flow, blade angle and operating efficiency; Dividing the water pump historical data set into a training set and a test set; Constructing a first support vector regression model and a second support vector regression model; Inputting the training set into the first support vector regression model, taking the lift and the flow rate as input and the blade angle as output, obtaining the optimal hyperplane by minimizing the objective function, so as to obtain a first initial support vector regression model; Inputting the training set into the second support vector regression model, taking the lift and the flow rate as input, taking the operating efficiency as output, and obtaining the optimal hyperplane by minimizing the objective function to obtain a second initial support vector regression model; The first initial support vector regression model and the second initial support vector regression model are tested and optimized using the test set to obtain two final support vector regression models that meet the first preset requirement; the two final support vector regression models constitute the preset support vector regression model.

4. The intelligent pump station characteristic correlation fitting optimization method according to claim 1 is characterized in that: The construction process of the preset BP neural network model includes: Acquire a water pump historical data set; each water pump historical data in the water pump historical data set includes head, flow, blade angle and operating efficiency; Dividing the water pump historical data set into a training set and a test set; Build BP network; Inputting the training set into the BP network, taking the lift and the flow rate as input, taking the blade angle and the operating efficiency as output, and performing network training according to a preset learning rate and a preset number of training times to obtain an initial BP neural network model; The initial BP neural network model is tested and optimized using the test set to obtain a final BP neural network model that meets the second preset requirement; the final BP neural network model is a preset BP neural network model.

5. The intelligent pump station characteristic correlation fitting optimization method according to claim 1 is characterized in that: The construction process of the preset random forest regression model includes: Acquire a water pump historical data set; each water pump historical data in the water pump historical data set includes head, flow, blade angle and operating efficiency; Extracting a plurality of water pump historical data as samples from the water pump historical data set with replacement to obtain a sample subset; For any sample in the sample subset, a binary decision tree is constructed using the CART algorithm; and the binary decision trees corresponding to multiple samples constitute a preset random forest regression model.

6. The intelligent pump station characteristic correlation fitting optimization method according to claim 1 is characterized in that: The preset support vector regression model, the preset BP neural network model and the preset random forest regression model are expressed as: ； in, To preset the support vector regression model, To preset the random forest regression model, It is the preset BP network model; is the output feature, To refer to or , To preset the number of binary decision trees in the random forest regression model, is the i-th binary decision tree in the preset random forest regression model, are the input characteristics, namely head and flow rate.

7. The intelligent pump station characteristic correlation fitting optimization method according to claim 2 is characterized in that: The preset decision model includes an input layer, a first fully connected layer, a second fully connected layer, an activation layer and an output layer which are arranged in sequence.

8. The intelligent pump station characteristic correlation fitting optimization method according to claim 4 is characterized in that: Inputting the training set into the BP network includes: normalizing each water pump historical data in the training set and then inputting the data into the BP network.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the intelligent pump station characteristic correlation fitting optimization method described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the intelligent pump station characteristic correlation fitting optimization method described in any one of claims 1 to 8 is implemented.

Images