Method for establishing electricity-to-water coefficient prediction model

Abstract

The invention relates to the field of water resource management, in particular to a method for establishing an electricity-to-water coefficient prediction model, which comprises the following steps: step S100, acquiring an electricity-to-water coefficient, water consumption and an influence factor data set of the electricity-to-water coefficient obtained by field test; step S200, performing correlation analysis on the electricity-to-water coefficient and the influence factor thereof by using a Pearson correlation analysis method to obtain a high-correlation influence factor data set; and step S300, preprocessing the high-correlation impact factor data set, and then inputting the high-correlation impact factor data set into a convolutional neural network prediction model for training to obtain an electricity-to-water coefficient prediction model with relatively high prediction precision. According to the method, the Pearson correlation analysis and the convolutional neural network are combined, the prediction precision of the electricity-to-water coefficient can be remarkably improved, the method has high generalization ability, high prediction accuracy can be kept in different regions and under different conditions, the prediction efficiency is greatly improved, and the method is suitable for popularization and application. The prediction model provided by the invention can efficiently and accurately predict the coefficient of water conversion by electricity.

Description

Technical Field

[0001] The invention relates to the technical field of water resource management, in particular to a method for establishing a prediction model based on electricity-to-water coefficient. Background Art

[0003] The basic principle of electricity-to-water conversion is to indirectly estimate the water volume of the pump by measuring the electricity consumption of the pump. Specifically, the actual water consumption can be estimated by using the ratio between the power consumption of the pump and the amount of water it outputs (i.e., the electricity-to-water conversion coefficient). The main advantage of electricity-to-water conversion is that it does not require the installation of complex flow meters or other expensive water volume monitoring equipment. The estimated value of water consumption can be obtained by relying solely on electricity data, which has significant cost advantages and practicality. In the fields of agricultural irrigation and groundwater exploitation, electricity-to-water conversion technology is widely used in groundwater monitoring and management, and can provide reliable data support and decision-making basis for the rational use and management of water resources. The determination and analysis of the electricity-to-water conversion coefficient can help the government and relevant departments to scientifically allocate water resources and avoid resource waste and environmental degradation caused by excessive exploitation.

[0004] Although the electricity-to-water technology has been applied to a certain extent in China, the overall research and technological development is still in the exploratory stage, and there are mainly the following challenges:

[0005] (1) There are many factors that affect the electricity-to-water conversion coefficient, including the type and efficiency of the water pump, groundwater level, irrigation method, grid voltage fluctuation, etc. There is a complex relationship between these factors. Some areas lack a complete information infrastructure, making data collection difficult.

[0006] (2) Although many existing studies use machine learning methods to predict the water-to-energy coefficient, most methods still have the problem of weak model generalization ability. This is because in practical applications, machine learning models have poor robustness and lack representativeness. Summary of the invention

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for establishing a prediction model based on electricity-to-water coefficient, comprising:

[0009] Step S100, obtaining a data set of electricity-to-water conversion coefficient, water consumption, and influencing factors of the electricity-to-water conversion coefficient obtained by field testing;

[0010] Step S200, using the Pearson correlation analysis method to perform correlation analysis on the electricity-to-water coefficient and its influencing factors, to obtain a high-correlation influencing factor data set;

[0011] Step S300, preprocessing the highly correlated influencing factor data set, and then inputting it into the convolutional neural network prediction model for training, so as to obtain a water-to-electricity coefficient prediction model with high prediction accuracy.

[0012] As a further solution of the present invention: the data set of factors affecting the electricity-to-water conversion coefficient, water consumption and the electricity-to-water conversion coefficient obtained by field testing specifically includes the following steps: selecting a number of representative typical wells in the study area as monitoring points to ensure that they are evenly distributed and cover different geological conditions, groundwater level depths and irrigation methods; installing monitoring equipment in typical wells to measure the electricity consumption, water extraction and other factors that may affect the electricity-to-water conversion coefficient of the typical wells; calculating the electricity-to-water conversion coefficient (defined as the ratio of water extraction to electricity consumption) through a formula, and annotating each sample data to form a complete electricity-to-water conversion coefficient data set.

[0013] As a further solution of the present invention: the data set of influencing factors of the electricity-to-water coefficient includes input data and output data, the input data includes irrigation method, groundwater depth, stable drawdown, well age, well depth, water pump power, water pump head and water pump flow; the output data includes the electricity-to-water coefficient.

[0014] As a further solution of the present invention: using Pearson correlation analysis to perform correlation analysis on the water-discount coefficient and its influencing factors, obtaining a high-correlation influencing factor data set specifically includes the following steps: extracting all input data and their corresponding water-discount coefficients to form a data matrix; using the Pearson correlation analysis method to calculate the Pearson correlation coefficient between each input data and the water-discount coefficient, and recording the correlation strength of each factor; based on the absolute value of the Pearson correlation coefficient, screening out influencing factors that are highly correlated with the water-discount coefficient. This process can effectively eliminate redundant data and retain those key factors that have important contributions to model prediction; pairing the screened high-correlation factors with the corresponding water-discount coefficients to construct a high-correlation influencing factor data set as input data for subsequent model training.

[0015] As a further solution of the present invention: preprocessing the high correlation influencing factor data set specifically includes data cleaning and data normalization. Data cleaning is to regard the situation where the output data cannot be generated or the generated data is unreasonable due to unreasonable input data design as data missing, and these data missing data are found and the group of data is deleted.

[0016] As a further solution of the present invention: data normalization is to perform normalization preprocessing on each group of initial data groups in turn. This is because the difference in order of magnitude will cause the attribute with a larger magnitude to dominate. The specific process is that for each attribute, minA and maxA are set as the maximum and minimum values ​​of attribute A respectively, and the original value x in A is mapped to the value x' in the interval [0,1] through normalization. The specific formula is as follows:

[0017] The normalized high-correlation influencing factor data and its corresponding electricity-to-water coefficient are stored in the data set as a complete set of data for subsequent model training.

[0018] As a further solution of the present invention: the information of the prediction model of the electric water coefficient is as follows: 70% of the data set is used as a training set for training the model, and the remaining 30% of the data set is used as a test set for testing the actual learning ability of the network; the performance and accuracy of the prediction network are measured by using the determination coefficient, loss function and evaluation function respectively. The determination coefficient is an indicator used to evaluate the prediction effect of the regression model, which represents the ratio of the degree of fit between the predicted value and the true value, and its definition is as follows:

[0019]

[0020] Where n represents the number of samples, Y i is the true value, is the predicted value, is the average of the true values. 2 The closer it is to 1, the better the prediction effect is.

[0021] The loss function is used to represent the difference between the true value and the predicted value. The convolutional neural network continuously updates its own model parameters by calculating the loss function during the training process. Therefore, the smaller the loss function, the better. The evaluation function is used to represent the degree of matching between the true value and the predicted value. The convolutional neural network ultimately uses the evaluation function to show the prediction effect of the network. The evaluation function is similar to the loss function, except that the result of the evaluation function is not used in the training process. The mean square error (MSE) is used as the loss function in the present invention, which is defined as follows:

[0022]

[0023] Where n represents the number of samples, Y i represents the true value, Represents the predicted value.

[0024] The input of the prediction network is the highly correlated electricity-to-water influencing factor obtained by Pearson correlation analysis, and the output of the prediction network is the electricity-to-water coefficient. The prediction network includes a convolutional layer of the first layer, three convolutional layers and three maximum pooling layers connected alternately, a flattening layer and three fully connected layers; zero padding uses the same mode, the activation function is the ReLu function, and the three pooling windows are (2, 1); the number of neurons in the three fully connected layers is 16, 16, and 16, and the activation functions of the three fully connected layers are all linear functions; the optimizer selects the sgdm mode.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] The present invention can significantly improve the accuracy of the prediction of the electricity-to-water coefficient by combining Pearson correlation analysis and convolutional neural network. Through Pearson correlation analysis, the influencing factors that are highly correlated with the electricity-to-water coefficient are screened out, redundant data is reduced and model efficiency is improved. The convolutional neural network effectively handles the complex nonlinear relationship between the influencing factors, solving the challenges that traditional methods cannot cope with. In addition, the method has a strong generalization ability, can maintain a high prediction accuracy in different regions and conditions, and greatly improves the prediction efficiency. The prediction model constructed using the present invention can efficiently and accurately predict the electricity-to-water coefficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 The figure is a flow chart of a method for establishing a prediction model based on the electricity-to-water coefficient in an embodiment of the present invention.

[0028] Figure 2 Detailed step diagram of S200 in the method for establishing a prediction model based on the electricity-to-water coefficient in an embodiment of the present invention.

[0029] Figure 3 This is a structural diagram of a convolutional neural network in a method for establishing a prediction model based on the electrical water deduction coefficient in an embodiment of the present invention.

[0030] Figure 4 This is a diagram showing the prediction effect of establishing a prediction model based on the electricity-to-water coefficient in an embodiment of the present invention.

[0031] Figure 5 The figure is a prediction error distribution diagram of the electricity-to-water coefficient prediction model established in the embodiment of the present invention. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] The present invention is a method for establishing a prediction model for the water-to-electricity coefficient, which can not only help us accurately screen out the influencing factors that are highly correlated with the water-to-electricity coefficient, such as water pump performance, groundwater level, etc., but also effectively reduce redundant features and build a more streamlined and efficient data set. In addition, this method also uses a convolutional neural network to learn the high-correlation influencing factor data set screened out by the Pearson correlation analysis, so that the network can automatically identify and extract the deep-level features between these factors. Through training, the model can obtain higher prediction accuracy while enhancing its generalization ability, thereby ensuring that the prediction effect of the method in different regions is more stable and reliable.

[0034] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0035] Example 1, see Figure 1-Figure 5 In 2024, the applicant conducted a test on the electricity-to-water coefficient at 281 representative wells in Baicheng City, Jilin Province. In this process, the input data affecting the electricity-to-water coefficient were collected, including eight parameters such as irrigation method, groundwater depth, drawdown when pumping is stable, well age, well depth, pump power, pump head, and pump rated flow, and the measured electricity-to-water coefficient was used as the output data.

[0036] Data correlation analysis: Through Pearson correlation analysis, the correlation coefficient between each input data and the electricity-to-water coefficient is calculated. The definition of the Pearson correlation coefficient is as follows:

[0037]

[0038] Wherein ρ represents the Pearson correlation coefficient, n represents the number of samples, x represents the input data, and z represents the output data. The value range of the Pearson correlation coefficient is [-1,1]. The closer its absolute value is to 1, the stronger the correlation is. The Pearson correlation is graded as follows: strong correlation (|ρ|≥0.5); moderate correlation (0.3≤|ρ|<0.50); ​​weak correlation (|ρ|<0.3). In the present invention, the calculation of the Pearson correlation coefficient is implemented by MATLAB software, and the final results are shown in Table 1.

[0039] Table 1 Pearson correlation analysis results

[0040] Input Data Pearson coefficient Correlation strength Groundwater depth 0.54 Strong correlation Irrigation method 0.50 Strong correlation Steady Descent 0.45 Moderately related Pump head 0.18 Weak correlation Water pump power 0.17 Weak correlation Well Age 0.13 Weak correlation Rated flow of water pump 0.11 Weak correlation well deep 0.08 Weak correlation

[0041] According to the classification of the results in Table 1, strongly correlated and moderately correlated parameters (irrigation method, groundwater depth and stable drawdown) were selected as the input data of the convolutional neural network model, and the electricity-to-water coefficient was used as the output data.

[0042] Convolutional neural network construction: According to the technical solution of the present invention, a convolutional neural network is used for modeling. The construction of the convolutional neural network model in the present invention is implemented by MATLAB software. The structure of the model includes: input layer, 3 convolutional layers, 3 maximum pooling layers, flattening layer and 3 fully connected layers. The activation function is the ReLU function, and the same filling mode is adopted. The collected data is divided into training set and test set in proportion, of which 70% of the data is used for training the model and 30% of the data is used for testing the actual prediction ability of the model. Through training, the R of the final model is 2 The value is 0.82, and the mean square error (MSE) is 0.78, indicating that the model has a high prediction effect on the electricity-to-water coefficient. Figure 4 As shown, the prediction error distribution is Figure 5 shown.

[0043] The actual application results are as follows: the prediction model was applied to the prediction of electricity-to-water coefficient in Baicheng City, Jilin Province. Figure 4 and Figure 5 The results show that the maximum prediction error between the predicted value and the measured value of the model is 28.79%, the minimum prediction error is 0.75%, and the average prediction error is 18.57%. The above data show that the model obtained by the present invention can accurately predict the electricity-to-water coefficient of different well points, providing efficient and reliable data support for the scientific management of water resources and water use regulation of agricultural irrigation. Practical applications show that the model obtained by the present invention significantly improves the accuracy of groundwater resource management in Baicheng City, reduces the errors caused by complex nonlinear relationships in traditional methods, and helps decision-making departments to achieve more accurate water consumption regulation. Through practice in Baicheng City, Jilin Province, the method of establishing a prediction model for electricity-to-water coefficient proposed in the present invention not only verifies the feasibility and effectiveness of its technical solution, but also demonstrates good generalization ability in different regions and conditions, providing a reliable solution for the prediction of electricity-to-water coefficient.

[0044] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A method for establishing a prediction model based on electricity-to-water coefficient, characterized in that: The following steps are involved: Step S100, obtaining a data set of electricity-to-water conversion coefficient, water consumption, and influencing factors of the electricity-to-water conversion coefficient obtained by field testing; Step S200, using the Pearson correlation analysis method to perform correlation analysis on the electricity-to-water coefficient and its influencing factors, to obtain a high-correlation influencing factor data set; Step S300, preprocessing the highly correlated influencing factor data set, and then inputting it into the convolutional neural network prediction model for training, so as to obtain a prediction model of the electricity-to-water coefficient with high prediction accuracy.

2. The method for establishing a prediction model based on electricity-to-water coefficient according to claim 1, characterized in that: In step S100, the data set of the electricity-to-water conversion coefficient, water consumption and influencing factors of the electricity-to-water conversion coefficient obtained by field testing is obtained by: selecting a number of representative typical wells in the study area as monitoring points to ensure that they are evenly distributed and cover different geological conditions, groundwater level depths and irrigation methods; installing monitoring equipment in the typical wells to measure the electricity consumption, water extraction and other factors that may affect the electricity-to-water conversion coefficient of the typical wells; calculating the electricity-to-water conversion coefficient by a formula, and annotating each sample data to form a complete data set of influencing factors of the electricity-to-water conversion coefficient.

3. The method for establishing a prediction model based on electricity-to-water coefficient according to claim 1 or 2, characterized in that: In step S100, the data set of influencing factors of the electricity-to-water coefficient includes input data and output data, the input data includes irrigation method, groundwater depth, stable drawdown, well age, well depth, water pump power, water pump head and water pump flow; the output data includes the electricity-to-water coefficient.

4. The method for establishing a prediction model based on electricity-to-water coefficient according to claim 3, characterized in that: In step 200, the correlation analysis of the water-discount coefficient and its influencing factors is performed using Pearson correlation analysis, and obtaining a high-correlation influencing factor data set specifically includes the following steps: extracting all input data and their corresponding water-discount coefficients to form a data matrix; using the Pearson correlation analysis method, respectively calculating the Pearson correlation coefficient between each input data and the water-discount coefficient, and recording the correlation strength of each factor; based on the absolute value of the Pearson correlation coefficient, screening out influencing factors that are highly correlated with the water-discount coefficient; pairing the screened high-correlation factors with the corresponding water-discount coefficients to construct a high-correlation influencing factor data set.

5. The method for establishing a prediction model based on electricity-to-water coefficient according to claim 1, characterized in that: The preprocessing of the high correlation influencing factor data set in step S300 specifically includes data cleaning and data normalization.

6. The method for establishing a prediction model based on electricity-to-water coefficient according to claim 5, characterized in that: The data cleaning is to regard the situation that the output data cannot be generated or the generated data is unreasonable due to unreasonable input data design as data missing, find these missing data and delete the group of data; data normalization is to perform normalization preprocessing on each group of initial data in turn. The specific process is that for each attribute, minA and maxA are set as the maximum and minimum values ​​of attribute A respectively, and the original value x in A is mapped to the value x' in the interval [0,1] through normalization. The specific formula is as follows:

7. The method for establishing a prediction model based on electricity-to-water coefficient according to claim 1, characterized in that: In the step S300, the structure of the electric water deduction coefficient prediction model includes a convolution layer, a pooling layer, a flattening layer and a fully connected layer. The first layer of the electric water deduction coefficient prediction model is the convolution layer, followed by three convolution layers and three maximum pooling layers that are alternately connected. The convolution layer adopts a zero-filling method, the mode is same, and the pooling window size is (2,1). After the convolution layer and the pooling layer, a flattening layer is added to map the high-dimensional features into a one-dimensional vector. The fully connected layer includes three layers, and the number of neurons is 16, 16 and 16 respectively. The activation functions of the fully connected layers are all linear functions. All convolution layers and pooling layers use ReLU activation functions, and the network optimization adopts the sgdm optimizer.

Images