Substation metering system state prediction method and device and electronic equipment

Abstract

The embodiment of the invention provides a substation metering system state prediction method and device and electronic equipment, and relates to the technical field of sampling algorithms. The method comprises the following steps: acquiring historical data of a transformer substation, wherein the historical data comprises an environment state quantity and a metering quantity corresponding to the environment state quantity; the historical data is preprocessed; processing the preprocessed data through a closed-loop clustering algorithm to obtain data classification in different scenes, and dividing the preprocessed data after data classification into a training set and a verification set; training a combined model by adopting the training set, and verifying the trained combined model by adopting the verification set; the combined model comprises a random forest algorithm, a convolutional neural network algorithm and an echo state network algorithm; the verified combined model is used for outputting a corresponding prediction result based on the input environment state quantity and the corresponding metering quantity. According to the implementation mode provided by the invention, the intelligent level of operation and maintenance of the transformer substation is effectively improved.

Description

Technical Field

[0001] The present application relates to the technical field of sampling algorithms, and specifically to a method for predicting the state of a substation metering system, a device for predicting the state of a substation metering system, an electronic device, a storage medium, and a computer program product. Background Art

[0002] With the development of intelligent power system, substation is an important part of power system, and its real-time monitoring and evaluation of operation status is crucial for the safe and stable operation of power grid. Traditional substation monitoring methods mainly rely on manual regular inspections and simple monitoring equipment, which have problems such as long detection cycle, data lag, and insufficient evaluation accuracy, which can hardly meet the needs of modern power grid for efficient and accurate monitoring.

[0003] During the operation of a substation, various environmental factors and equipment status will have a significant impact on the stability of the system, such as temperature, humidity, load conditions, equipment aging, etc. Real-time acquisition and analysis of these environmental status quantities and metering data are of great significance for preventing equipment failures and optimizing operation and maintenance decisions. However, the substation site environment is complex, the data sources are diverse, and the data volume is huge. Traditional monitoring methods are difficult to achieve efficient data processing and real-time status evaluation. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide a method, device and electronic device for predicting the status of a substation metering system, combined with advanced deep learning algorithms, to establish a prediction model and real-time evaluation system for the equipment status, to achieve real-time monitoring, prediction and evaluation of the substation equipment status, so as to at least solve some of the problems in the background technology.

[0005] In order to achieve the above-mentioned purpose, a method for predicting the state of a substation metering system is provided in the present application, which method includes: obtaining historical data of the substation including environmental state quantities and corresponding metering quantities; preprocessing the historical data to obtain preprocessed data; processing the preprocessed data through a closed-loop clustering algorithm to obtain data classification under different scenarios, and dividing the preprocessed data after data classification into a training set and a verification set; inputting the data in the training set into a random forest algorithm to obtain the corresponding output, and correspondingly fusing the input data and output data of the random forest algorithm to obtain a processed training set; using the processed training set to train a combined model, and using the verification set to verify the trained combined model; the combined model includes a convolutional neural network part and an echo state network part; the verified combined model is used to output a corresponding prediction result based on the input environmental state quantity and the corresponding metering quantity.

[0006] Optionally, historical data of the substation including environmental state quantities and corresponding metering quantities are obtained, including: collecting historical data of the substation, the historical data including environmental state quantities and corresponding metering quantities; distinguishing the environmental state quantities and the corresponding metering quantities based on different scenarios, and combining the environmental state quantities and metering quantities with corresponding relationships into a state matrix form.

[0007] Optionally, the historical data is preprocessed to obtain preprocessed data, including: constructing an equipment state matrix based on the historical data; performing numerical interval judgment on the elements in the equipment state matrix, and determining the values ​​that are not in the preset numerical interval as abnormal values; after deleting the abnormal values, using the Lagrange interpolation method to generate new values ​​to supplement the positions of the deleted abnormal values; standardizing the deviations of the elements in the equipment state matrix through linear transformation; and using the principal component analysis method to perform dimensionality reduction processing on the equipment state matrix after the deviation standardization to obtain the preprocessed data.

[0008] Optionally, the preprocessed data is processed by a closed-loop clustering algorithm to obtain data classifications for different scenarios, including: inputting the preprocessed data into a closed-loop clustering algorithm model to divide all scenarios into several categories; based on the measurement quantities of the scenarios in the same category after clustering as the prediction objects, respectively training convolutional neural network models with the goal of minimizing the root mean square error to obtain prediction models that correspond one to one with the categories after clustering; the data of each scenario in all scenarios are processed as follows: respectively using the prediction models for prediction to obtain multiple groups of prediction results equal to the number of prediction models; if a group of prediction results among the multiple groups of prediction results has the smallest root mean square error, then moving the scene corresponding to the group of prediction results to the corresponding category; repeating the aforementioned steps of obtaining the prediction model and the data processing steps for each scenario until the end condition is reached, the end condition including that the clustering result no longer changes or the maximum number of iterations is reached; and taking the clustering result when the end condition is reached as the data classification for the different scenarios.

[0009] Optionally, the method further includes: generating clustering labels according to the clustering results when the termination condition is reached and adding them to the preprocessed data; and performing imbalance processing on the preprocessed data after data classification in different scenarios.

[0010] Optionally, the random forest algorithm in the combined model is configured as follows: the data in the training set is input into the random forest algorithm to obtain corresponding outputs, and the input data and output data of the random forest algorithm are correspondingly fused to obtain fused data; the convolutional neural network algorithm in the combined model is configured as follows: a feature extraction result is obtained based on the input fused data; the echo state network algorithm in the combined model is configured as follows: a state matrix is ​​generated based on the feature extraction results, and the state matrix obtains the final output result through the output weight matrix.

[0011] Optionally, the input data and output data of the random forest algorithm are correspondingly fused to obtain fused data, including: the input data and output data of the random forest algorithm are spliced ​​to obtain fused data; the echo state network algorithm is trained by the following steps: initializing the weight matrix and hyperparameters in the echo state network algorithm; the data in the training set is passed through the convolutional neural network part to generate corresponding feature extraction results, the feature extraction results are input into the echo state network part, and the connection weights from the hidden layer to the output layer are trained; the hidden layer obtains the final output result through the output weight matrix of the output layer.

[0012] Optionally, the verified combined model is used to output a corresponding prediction result based on the input environmental state quantity and the corresponding metering quantity, including: obtaining real-time data of the environmental state quantity and the corresponding metering quantity; constructing the real-time data in the form of a state matrix; inputting the real-time data in the form of the state matrix into the verified combined model, and taking the output of the verified combined model as the corresponding prediction result.

[0013] Optionally, after outputting a prediction result corresponding to the input environmental state quantity and the corresponding metering quantity, the method further includes: fitting the prediction result output by the verified combined model into a dynamic prediction curve; calculating the residual between the dynamic prediction curve and the actual curve obtained by fitting the real-time data within a set time interval, and obtaining a residual dynamic change curve based on the residual; and determining the equipment status of the substation according to the fluctuation condition of the residual dynamic change curve.

[0014] The present application also provides a device for predicting the state of a substation metering system, the device comprising:

[0015] A data acquisition module is used to acquire historical data of the substation including environmental state quantities and corresponding metering quantities; a preprocessing module is used to preprocess the historical data to obtain preprocessed data; a data clustering module is used to process the preprocessed data through a closed-loop clustering algorithm to obtain data classification under different scenarios, and divide the preprocessed data after data classification into a training set and a verification set; and a combination model module is used to train the combination model using the training set and verify the trained combination model using the verification set; the combination model includes a random forest algorithm, a convolutional neural network algorithm and an echo state network algorithm; the verified combination model is used to output a corresponding prediction result based on the input environmental state quantity and the corresponding metering quantity.

[0016] Optionally, historical data of the substation including environmental state quantities and corresponding metering quantities are obtained, including: collecting historical data of the substation, the historical data including environmental state quantities and corresponding metering quantities; distinguishing the environmental state quantities and the corresponding metering quantities based on different scenarios, and combining the environmental state quantities and metering quantities with corresponding relationships into a state matrix form.

[0017] Optionally, the historical data is preprocessed to obtain preprocessed data, including: constructing an equipment state matrix based on the historical data; performing numerical interval judgment on the elements in the equipment state matrix, and determining the values ​​that are not in the preset numerical interval as abnormal values; after deleting the abnormal values, using the Lagrange interpolation method to generate new values ​​to supplement the positions of the deleted abnormal values; standardizing the deviations of the elements in the equipment state matrix through linear transformation; and using the principal component analysis method to perform dimensionality reduction processing on the equipment state matrix after the deviation standardization to obtain the preprocessed data.

[0018] Optionally, the preprocessed data is processed by a closed-loop clustering algorithm to obtain data classifications for different scenarios, including: inputting the preprocessed data into a closed-loop clustering algorithm model to divide all scenarios into several categories; based on the measurement quantities of the scenarios in the same category after clustering as the prediction objects, respectively training convolutional neural network models with the goal of minimizing the root mean square error to obtain prediction models that correspond one to one with the categories after clustering; the data of each scenario in all scenarios are processed as follows: respectively using the prediction models for prediction to obtain multiple groups of prediction results equal to the number of prediction models; if a group of prediction results among the multiple groups of prediction results has the smallest root mean square error, then moving the scene corresponding to the group of prediction results to the corresponding category; repeating the aforementioned steps of obtaining the prediction model and the data processing steps for each scenario until the end condition is reached, the end condition including that the clustering result no longer changes or the maximum number of iterations is reached; and taking the clustering result when the end condition is reached as the data classification for the different scenarios.

[0019] Optionally, the device further comprises an imbalance processing module, which is used to: generate clustering labels according to the clustering results when the termination condition is reached and add them to the preprocessed data; and perform imbalance processing on the preprocessed data after data classification in different scenarios.

[0020] Optionally, the random forest algorithm in the combined model is configured as follows: the data in the training set is input into the random forest algorithm to obtain corresponding outputs, and the input data and output data of the random forest algorithm are correspondingly fused to obtain fused data; the convolutional neural network algorithm in the combined model is configured as follows: a feature extraction result is obtained based on the input fused data; the echo state network algorithm in the combined model is configured as follows: a state matrix is ​​generated based on the feature extraction results, and the state matrix obtains the final output result through the output weight matrix.

[0021] Optionally, the input data and output data of the random forest algorithm are correspondingly fused to obtain fused data, including: the input data and output data of the random forest algorithm are spliced ​​to obtain fused data; the echo state network algorithm is trained by the following steps: initializing the weight matrix and hyperparameters in the echo state network algorithm; the data in the training set is passed through the convolutional neural network part to generate corresponding feature extraction results, the feature extraction results are input into the echo state network part, and the connection weights from the hidden layer to the output layer are trained; the hidden layer obtains the final output result through the output weight matrix of the output layer.

[0022] Optionally, the verified combined model is used to output a corresponding prediction result based on the input environmental state quantity and the corresponding metering quantity, including: obtaining real-time data of the environmental state quantity and the corresponding metering quantity; constructing the real-time data in the form of a state matrix; inputting the real-time data in the form of the state matrix into the verified combined model, and taking the output of the verified combined model as the corresponding prediction result.

[0023] Optionally, the device also includes a state determination module, which is used to: after outputting a prediction result corresponding to the input environmental state quantity and the corresponding metering quantity, fit the prediction result output by the verified combined model into a dynamic prediction curve; calculate the residual between the dynamic prediction curve and the actual curve obtained by fitting the real-time data within a set time interval, and obtain a residual dynamic change curve based on the residual; determine the equipment state of the substation according to the fluctuation condition of the residual dynamic change curve.

[0024] The present application also provides an electronic device, comprising: at least one processor; a memory connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the at least one processor implements the aforementioned substation metering system status prediction method by executing the instructions stored in the memory.

[0025] The present application also provides a machine-readable storage medium, on which instructions are stored. When the instructions are executed by a processor, the processor is configured to execute the aforementioned method for predicting the state of a substation metering system.

[0026] The present application also provides a computer program product, including a computer program, which implements the aforementioned substation metering system state prediction method when executed by a processor.

[0027] The above technical solution has the following beneficial effects:

[0028] (1) Through the closed-loop clustering algorithm and random forest algorithm, the accuracy and robustness of the clustering results are significantly improved, thereby improving the stability of the subsequent model.

[0029] (2) The model of random forest combined with CNN-ESN uses CNN to extract high-dimensional features, ESN to process time series data, and random forest to improve model stability. It can show higher prediction accuracy and robustness in complex data and sequence prediction tasks;

[0030] (3) It can effectively improve the intelligence level of substation operation and maintenance, enhance the safety and reliability of power grid operation, and has important application value and promotion prospects.

[0031] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present application, but do not constitute a limitation on the embodiments of the present application. In the accompanying drawings:

[0033] Figure 1 A schematic diagram of the steps of a method for predicting the state of a substation metering system according to an embodiment of the present application is schematically shown;

[0034] Figure 2 A schematic diagram of a technical route of a method for predicting a state of a substation metering system according to an embodiment of the present application is schematically shown;

[0035] Figure 3 The schematic diagram of the implementation of the data preprocessing step in the method for predicting the state of a substation metering system according to the embodiment of the present application is schematically shown;

[0036] Figure 4 The following schematically shows an implementation diagram of a closed-loop clustering algorithm in a method for predicting a state of a substation metering system according to an embodiment of the present application;

[0037] Figure 5Schematically shows a data processing schematic diagram of a combined model in a method for predicting a state of a substation metering system according to an embodiment of the present application;

[0038] Figure 6 The schematic diagram of the structure of the device for predicting the state of a metering system of a substation according to the embodiment of the present application is shown;

[0039] Figure 7 The internal structure of an electronic device according to an embodiment of the present application is schematically shown. DETAILED DESCRIPTION

[0040] The specific implementation of the embodiment of the present application is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present application, and is not used to limit the embodiment of the present application.

[0041] Figure 1 The following is a schematic diagram showing the steps of a method for predicting the state of a substation metering system according to an embodiment of the present application. Figure 1 As shown, the method includes:

[0042] S01. Obtain historical data of the substation including environmental status quantities and corresponding metering quantities;

[0043] S02, preprocessing the historical data to obtain preprocessed data;

[0044] S03, processing the preprocessed data by a closed-loop clustering algorithm to obtain data classification under different scenarios, and dividing the preprocessed data after data classification into a training set and a validation set;

[0045] S04. The training set is used to train the combined model, and the verification set is used to verify the trained combined model; the combined model includes a random forest algorithm, a convolutional neural network algorithm and an echo state network algorithm; the verified combined model is used to output a prediction result corresponding to the input environmental state quantity and the corresponding metering quantity.

[0046] In the above implementation, steps S01 to S04 are mainly implemented in the model construction scenario. The model use step in this application is mainly to deploy the verified combined model online and use it to output the corresponding prediction results based on the input environmental state quantity and the corresponding metering quantity.

[0047] Through the above implementation methods, different scenarios of training data are effectively classified through data preprocessing and closed-loop clustering algorithms, which improves the scenario targeting of the training process. The data is classified and processed through the random forest algorithm, which improves the stability of the model. At the same time, the CNN-ESN model composed of convolutional neural networks and echo state networks extracts high-dimensional features through CNN (Convolutional Neural Networks) and processes time series data through ESN (Echo State Networks). It can show higher prediction accuracy and robustness in complex data and sequence prediction tasks. The model improves the accuracy of the prediction results. Through the application of the above multiple technologies in substation equipment monitoring and maintenance, the intelligent level of substation operation and maintenance can be effectively improved, and the safety and reliability of power grid operation can be improved, which has important application value and promotion prospects.

[0048] Figure 2 The following schematically shows a technical route diagram of the method for predicting the state of a substation metering system according to the embodiment of the present application. Figure 2 As shown, the substation metering system state prediction method in this embodiment mainly includes: data collection step, data preprocessing step, cluster analysis step, unbalanced data processing step, random forest-neural network algorithm training, residual calculation step and implementation evaluation step.

[0049] In some embodiments of the present application, obtaining historical data of a substation including an environmental state quantity and a corresponding metering quantity includes: collecting historical data of the substation, the historical data including the environmental state quantity and the corresponding metering quantity; distinguishing the environmental state quantity and the corresponding metering quantity by different scenarios, and combining the environmental state quantity and the metering quantity with corresponding relationship into a state matrix form. For example, collecting environmental state quantities and metering quantities of a substation, distinguishing by different scenarios, collecting environmental state quantities and metering quantities under each scenario, and combining them into a state matrix form, denoted as M, assuming that there are k groups of scenarios and corresponding variables.

[0050] In some implementations of the present application, the historical data is preprocessed to obtain preprocessed data. Figure 3 The following is a schematic diagram showing the implementation of the data preprocessing step in the method for predicting the state of a substation metering system according to the embodiment of the present application. Figure 3 As shown, it includes: constructing a device state matrix based on the historical data. For example, to establish a device state matrix, the environmental state quantity and the measurement quantity considering the time series are respectively recorded as a i,p and b j,p, where a represents the environmental state quantity, which is i in total; b represents the measurement quantity, which is j in total, i and j are in the range of [0, k], and p represents different time nodes in a day. To ensure the reasonable composition of the device state matrix, the sampling nodes of the time in a day are consistent with the final number of scenes k. The final device state matrix structure is as follows:

[0051]

[0052] The final device state matrix M is a k-order square matrix. and represents the environmental state quantity and the measurement quantity, k is the number of scenes, i is the number of environmental state quantities, j is the number of measurement quantities, and p is the time series.

[0053] The elements in the device status matrix are judged by numerical intervals, and the values ​​that are not in the preset numerical interval are determined as abnormal values; for example: this embodiment adopts a simple statistical analysis method, and determines unreasonable data by performing descriptive analysis on the maximum and minimum values ​​of the data set, thereby screening and locating abnormal data, and the maximum and minimum values ​​are in a reasonable interval, which is determined by the actual situation of reality and equipment:

[0054]

[0055] Among them, max(a i ),min(a i ) are respectively represented as data a i The maximum and minimum values ​​of max(b i ),min(b i ) are respectively represented as data b i The maximum and minimum values ​​of .

[0056] After deleting the outlier, Lagrange interpolation is used to generate new values ​​to supplement the deleted outlier. The principle of Lagrange interpolation is to substitute the point corresponding to the missing data into the Lagrange interpolation formula to obtain the approximate data at that location. The elements in the device state matrix are normalized by linear transformation; for example:

[0057]

[0058] Among them, k represents the dimension of the data, a m 、a n 、b m 、b n Represents the data in the data set, and the left side of the equal sign is the supplementary value for the missing data.

[0059] The elements in the device state matrix are standardized by linear transformation, for example, data standardization is performed using the MAX-MIN standardization method. The MAX-MIN standardization is performed by repeating the linear transformation process of the data through a linear transformation method so that each eigenvalue is within [0, 1]. The solution of the MAX-MIN standardization is as follows:

[0060]

[0061] Where: max(a i )、min(a i )、max(b i )、min(b i ) represents the maximum and minimum values ​​of the data. After normalization, the data in the state matrix is ​​within [0,1], and the matrix is ​​recorded as M'.

[0062] The principal component analysis method is used to perform dimensionality reduction processing on the device state matrix after the deviation is standardized to obtain the pre-processed data. For example, the covariance matrix of the standardized data is calculated: M'M' T , then perform eigenvalue decomposition on the covariance matrix to obtain several eigenvalues, and then take the eigenvectors corresponding to the largest x eigenvalues ​​as the principal components, which explain the largest variance in the data. The selected x eigenvectors are combined into a dimensionality reduction matrix P, where each column is an eigenvector.

[0063] P=[v1,v2,...,v k ]

[0064] Multiply the dimension reduction matrix P by the standardized data M', project the data from the high-dimensional space to the low-dimensional space, and record the data in the standardized and dimension-reduced matrix as c i,j , the matrix is ​​recorded as N:

[0065] N=M'P.

[0066] The above method realizes data preprocessing. Compared with the problem that the data preprocessing technology in the prior art has limited processing capabilities for large amounts of data, the implementation method in this application further improves the data quality and feature utilization efficiency through abnormal data screening and feature dimensionality reduction, which helps to build a higher performance machine learning model. This improvement has obvious advantages for processing complex data sets, especially those containing noise and high-dimensional data.

[0067] In some embodiments of the present application, the preprocessed data is processed by a closed-loop clustering algorithm to obtain data classifications for different scenarios, including: inputting the preprocessed data into a closed-loop clustering algorithm model to divide all scenarios into several categories; based on the metering quantities of the scenarios in the same category after clustering as prediction objects, respectively training convolutional neural network models with the goal of minimizing the root mean square error to obtain prediction models that correspond one to one with the categories after clustering; the data of each scenario in all scenarios are processed as follows: respectively using the prediction models for prediction to obtain multiple groups of prediction results equal to the number of prediction models; if a group of prediction results among the multiple groups of prediction results has the smallest root mean square error, then moving the scene corresponding to the group of prediction results to the corresponding category; repeating the aforementioned steps of obtaining the prediction model and the data processing steps for each scenario until the end condition is reached, the end condition including that the clustering result no longer changes or the maximum number of iterations is reached; and taking the clustering result when the end condition is reached as the data classification for the different scenarios. Figure 4 The following schematically shows the implementation diagram of the closed-loop clustering algorithm in the method for predicting the state of the substation metering system according to the embodiment of the present application. Figure 4 As shown, for example, it can be implemented by the following steps:

[0068] Step a, initialization. Input the matrix N into the improved closed-loop K-means clustering algorithm model, and use the K-means algorithm to divide all k scenes in the area into v categories (1≤v≤k);

[0069] Step b: Prediction. Based on the clustering results, the measurement quantities of scenes in the same category are used as prediction objects, and convolutional neural network models are trained with the goal of minimizing the root mean squared error (RMSE) to obtain v prediction models corresponding to v categories:

[0070]

[0071] Among them, c i,j is the data in the feature matrix T, is the data average.

[0072] Step c: Feedback. For each scenario, v prediction models are applied to make predictions, and v groups of prediction results are obtained. If the prediction result of the rth group (1≤r≤v) has the smallest RMSE, the scenario is moved to the rth category. This process is repeated for all k scenarios to obtain new clustering results.

[0073] Step d: Iteration: Repeat steps b and c until the clustering result does not change or the maximum number of iterations is reached.

[0074] Through this implementation, a closed-loop clustering algorithm is used to classify data, which improves the stability of classification compared with the existing clustering algorithm.

[0075] In some embodiments of the present application, the method further includes: generating a clustering label according to the clustering result when the termination condition is reached and adding it to the preprocessed data; and performing unbalanced processing on the preprocessed data after data classification in different scenarios. After clustering is completed, the form of the device state matrix will not change much, and only a label will be generated in the last column of the matrix to represent each cluster after clustering. Then, unbalanced learning is performed on the matrix to solve the possible problem of uneven data distribution, and finally the new state matrix is ​​recorded as T.

[0076] The imbalanced processing provided by this implementation improves the prediction capability of the subsequent model for a small number of scenarios, thereby improving the performance of the overall model.

[0077] In some embodiments of the present application, the data in the training set is input into a random forest algorithm to obtain a corresponding output, and the input data and output data of the random forest algorithm are correspondingly fused to obtain a processed training set, including: after the data in the training set is input into the random forest algorithm, output data corresponding to the input data is obtained, and the output data is the predicted output of the input data; the input data and output data of the random forest algorithm are spliced ​​in a splicing manner to obtain a processed training set. For example: after clustering, the new state matrix T has been labeled according to the clustering results. After the data is divided into a training set T1 and a test set T2, the training set is input into the random forest algorithm. The predicted output of the random forest algorithm for the training data is obtained as the input feature of the neural network. These features are usually the prediction results of each tree, or the distribution of leaf nodes. After classification by the random forest algorithm, a new feature matrix T will be obtained. RF , combined with the input matrix, the resulting matrix T combined The next stage model will be:

[0078] T combined =[T1 T RF ]

[0079] Through the random forest algorithm processing in this implementation, the classification of training data is made more accurate and stable, and the stability of subsequent models is improved.

[0080] Figure 5 The following is a schematic diagram showing the data processing of the combined model in the method for predicting the state of the substation metering system according to the embodiment of the present application. Figure 5As shown, the training set is used to train the combined model, and the validation set is used to validate the trained combined model; the combined model includes a random forest algorithm, a convolutional neural network algorithm, and an echo state network algorithm. In some embodiments of the present application, the training of each part in the combined model includes: initializing the weight matrix and hyperparameters of the echo state network part in the combined model; initializing the CNN-ESN model includes: designing the convolution layer, pooling layer, etc. of CNN to extract the spatial features of the data; initializing ESN includes initializing the weight matrix of ESN, including input weights, echo state weights, and output weights. Determine the hidden layer size of ESN, that is, the number of echo state units, and hyperparameters such as connection sparsity. After the data in the processed training set passes through the convolutional neural network part, the corresponding feature extraction results are generated, and the feature extraction results are input into the echo state network part to train the connection weights from the hidden layer to the output layer; T combined Input the CNN-ESN model, train the CNN-ESN model, and extract high-dimensional features. The hidden layer obtains the final output result through the output weight matrix of the output layer. For example: After the input training data passes through CNN, a feature extraction result X is generated. cnn , send it to the ESN model, use the input features to update the dynamic state of ESN, and generate the state matrix:

[0081] h(t)=tanh(W in X cnn (t))+Wh(t-1))

[0082] Among them, X cnn (t) is the input feature, h(x) is the hidden state of ESN, and W in is the input weight, and W is the echo state weight. The state matrix h(t) is obtained by outputting the weight matrix W out Get the final output result Y:

[0083] Y=W out h(t)

[0084] Where Y is the true output.

[0085] After the training process is completed, the performance of the trained CNN-ESN model is evaluated and tuned using the validation set T2. The combined model after evaluation and tuning can be used for deployment applications, and after deployment, it is used to output the corresponding prediction results based on the input environmental state quantity and the corresponding metering quantity.

[0086] In this embodiment, the CNN-ESN model extracts high-dimensional features through CNN and processes time series data through ESN, which can show higher prediction accuracy and robustness in complex data and sequence prediction tasks.

[0087] In some embodiments of the present application, the verified combined model is used to output a corresponding prediction result based on the input environmental state quantity and the corresponding metering quantity, including: obtaining real-time data of the environmental state quantity and the corresponding metering quantity; constructing the real-time data in the form of a state matrix; inputting the real-time data in the form of the state matrix into the verified combined model, and taking the output of the verified combined model as the corresponding prediction result. When the combined model is deployed online after training and verification, it can obtain prediction results based on real-time data in the same form as the training data. This embodiment provides the deployment and application of a trained and verified combined model. The combined model is applied in the operation of the substation to obtain and analyze these environmental state quantities and metering data in real time. The prediction results output by the combined model are of great significance for preventing equipment failures and optimizing operation and maintenance decisions.

[0088] In some embodiments of the present application, after the prediction result corresponding to the input environmental state quantity and the corresponding metering quantity output is obtained, the method further includes: fitting the prediction result output by the verified combined model into a dynamic prediction curve; calculating the residual between the dynamic prediction curve and the actual curve obtained by fitting the real-time data within a set time interval, and obtaining the residual dynamic change curve based on the residual; determining the equipment state of the substation according to the fluctuation condition of the residual dynamic change curve. By means of dynamic curve fitting, the prediction output output by the CNN-ESN model is expressed as a dynamic prediction curve y(t), the actual curve is obtained by fitting the real-time data, and the residual between the prediction curve and the actual curve is calculated within a set time interval. After the residual is calculated, the residual dynamic change curve is drawn. When the amplitude of the residual dynamic curve frequently exceeds the set threshold within the set time interval, it is determined that there is a problem with the equipment state, thereby achieving real-time monitoring and early warning functions. This embodiment can realize automatic comparison with actual data based on the prediction output output by the CNN-ESN model, and realizes real-time monitoring, prediction and evaluation of the substation equipment state. This method can effectively improve the intelligence level of substation operation and maintenance, enhance the safety and reliability of power grid operation, and has important application value and promotion prospects.

[0089] Based on the same inventive concept, the present application also provides a device for predicting the state of a substation metering system. Figure 6 The schematic diagram shows the structure of the device for predicting the state of the substation metering system according to the embodiment of the present application. Figure 6As shown, the device includes: a data acquisition module, which is used to acquire historical data of the substation including environmental state quantities and corresponding metering quantities; a preprocessing module, which is used to preprocess the historical data to obtain preprocessed data; a data clustering module, which is used to process the preprocessed data through a closed-loop clustering algorithm to obtain data classification under different scenarios, and divide the preprocessed data after data classification into a training set and a verification set; and a combination model module, which is used to train the combination model using the training set and verify the trained combination model using the verification set; the combination model includes a random forest algorithm, a convolutional neural network algorithm and an echo state network algorithm; the verified combination model is used to output a corresponding prediction result based on the input environmental state quantity and the corresponding metering quantity.

[0090] In some optional embodiments of the present application, historical data of the substation including environmental state quantities and corresponding metering quantities are obtained, including: collecting historical data of the substation, the historical data including environmental state quantities and corresponding metering quantities; distinguishing the environmental state quantities and the corresponding metering quantities based on different scenarios, and combining the environmental state quantities and metering quantities with corresponding relationships into a state matrix form.

[0091] In some optional embodiments of the present application, the historical data is preprocessed to obtain preprocessed data, including: constructing a device state matrix based on the historical data; performing numerical interval judgment on the elements in the device state matrix, and determining the values ​​that are not in the preset numerical interval as abnormal values; after deleting the abnormal values, using the Lagrange interpolation method to generate new values ​​to supplement the positions of the deleted abnormal values; standardizing the deviations of the elements in the device state matrix through linear transformation; and using the principal component analysis method to perform dimensionality reduction processing on the device state matrix after the deviation standardization to obtain the preprocessed data.

[0092] In some optional embodiments of the present application, the preprocessed data is processed by a closed-loop clustering algorithm to obtain data classifications for different scenarios, including: inputting the preprocessed data into a closed-loop clustering algorithm model to divide all scenarios into several categories; based on the metering quantities of the scenes in the same category after clustering as the prediction objects, respectively training convolutional neural network models with the goal of minimizing the root mean square error to obtain prediction models that correspond one to one with the categories after clustering; the data of each scene in all scenarios are processed as follows: respectively using the prediction models for prediction to obtain multiple groups of prediction results equal to the number of prediction models; if a group of prediction results among the multiple groups of prediction results has the smallest root mean square error, then moving the scene corresponding to the group of prediction results to the corresponding category; repeating the aforementioned steps of obtaining the prediction model and the data processing steps for each scene until the end condition is reached, the end condition including that the clustering result no longer changes or the maximum number of iterations is reached; and taking the clustering result when the end condition is reached as the data classification for the different scenarios.

[0093] In some optional embodiments of the present application, the device also includes an imbalance processing module, which is used to: generate clustering labels based on the clustering results when the termination conditions are reached and add them to the preprocessed data; and perform imbalance processing on the preprocessed data after data classification in different scenarios.

[0094] In some optional embodiments of the present application, the random forest algorithm in the combined model is configured as follows: the data in the training set is input into the random forest algorithm to obtain the corresponding output, and the input data and output data of the random forest algorithm are correspondingly fused to obtain fused data; the convolutional neural network algorithm in the combined model is configured as follows: a feature extraction result is obtained based on the input fused data; the echo state network algorithm in the combined model is configured as follows: a state matrix is ​​generated based on the feature extraction result, and the state matrix obtains the final output result through the output weight matrix.

[0095] In some optional embodiments of the present application, the input data and output data of the random forest algorithm are correspondingly fused to obtain fused data, including: the input data and output data of the random forest algorithm are spliced ​​to obtain fused data; the echo state network algorithm is trained by the following steps: initializing the weight matrix and hyperparameters in the echo state network algorithm; the data in the training set is passed through the convolutional neural network part to generate corresponding feature extraction results, the feature extraction results are input into the echo state network part, and the connection weights from the hidden layer to the output layer are trained; the hidden layer obtains the final output result through the output weight matrix of the output layer.

[0096] In some optional embodiments of the present application, the verified combined model is used to output a corresponding prediction result based on the input environmental state quantity and the corresponding metered quantity, including: obtaining real-time data of the environmental state quantity and the corresponding metered quantity; constructing the real-time data in the form of a state matrix; inputting the real-time data in the form of the state matrix into the verified combined model, and using the output of the verified combined model as the corresponding prediction result.

[0097] In some optional embodiments of the present application, the device also includes a state determination module, which is used to: fit the prediction result output by the verified combined model into a dynamic prediction curve after outputting the prediction result based on the input environmental state quantity and the corresponding metering quantity; calculate the residual between the dynamic prediction curve and the actual curve obtained by fitting the real-time data within a set time interval, and obtain a residual dynamic change curve based on the residual; determine the equipment state of the substation according to the fluctuation condition of the residual dynamic change curve.

[0098] The specific definition of each functional module in the above-mentioned substation metering system state prediction device can refer to the definition of the substation metering system state prediction method above, which will not be repeated here. Each module in the above-mentioned system can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the electronic device in the form of hardware, or can be stored in the memory of the electronic device in the form of software, so that the processor can call and execute the operations corresponding to the above modules. It also achieves the advantages of improving the intelligence level of substation operation and maintenance and enhancing the safety and reliability of power grid operation.

[0099] In some embodiments of the present application, an electronic device is also provided, comprising: at least one processor; a memory connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the at least one processor executes the aforementioned substation metering system state prediction method. Its internal structure diagram can be as follows Figure 7 shown. Figure 7The internal structure diagram of an electronic device according to an embodiment of the present application is schematically shown. The electronic device includes a processor A01, a network interface A02, a memory (not shown in the figure) and a database (not shown in the figure) connected via a system bus. Among them, the processor A01 of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes an internal memory A03 and a non-volatile storage medium A04. The non-volatile storage medium A04 stores an operating system B01, a computer program B02 and a database (not shown in the figure). The internal memory A03 provides an environment for the operation of the operating system B01 and the computer program B02 in the non-volatile storage medium A04. The network interface A02 of the electronic device is used to communicate with an external terminal through a network connection. When the computer program B02 is executed by the processor A01, a method for predicting the state of a substation metering system is implemented.

[0100] Those skilled in the art will understand that Figure 7 The structure shown in the figure is merely a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the electronic device to which the scheme of the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0101] In one embodiment provided in the present application, a machine-readable storage medium is provided, on which instructions are stored. When the instructions are executed by a processor, the processor is configured to execute the aforementioned substation metering system state prediction method.

[0102] In one embodiment provided in the present application, a computer program product is provided, including a computer program, which implements the aforementioned substation metering system state prediction method when executed by a processor.

[0103] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0104] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0105] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0106] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0107] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0108] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0109] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0110] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0111] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for predicting the state of a substation metering system, characterized in that: The method includes: Obtaining historical data of the substation including environmental status quantities and corresponding metering quantities; Preprocessing the historical data to obtain preprocessed data; Processing the preprocessed data by a closed-loop clustering algorithm to obtain data classification under different scenarios, and dividing the preprocessed data after data classification into a training set and a validation set; The training set is used to train the combined model, and the verification set is used to verify the trained combined model; the combined model includes a random forest algorithm, a convolutional neural network algorithm and an echo state network algorithm; the verified combined model is used to output a corresponding prediction result based on the input environmental state quantity and the corresponding metering quantity.

2. The method according to claim 1, characterized in that Obtain historical data of the substation including environmental status quantities and corresponding metering quantities, including: Collecting historical data of the substation, wherein the historical data includes environmental state quantities and corresponding metering quantities; The environmental state quantities and their corresponding measurement quantities are distinguished according to different scenarios, and the environmental state quantities and measurement quantities with corresponding relationships are combined into a state matrix form.

3. The method according to claim 1, characterized in that Preprocessing the historical data to obtain preprocessed data includes: constructing a device status matrix based on the historical data; Performing value interval judgment on the elements in the device state matrix, and determining the values ​​that are not in the preset value interval as abnormal values; After deleting the outlier value, a new value is generated by using the Lagrange interpolation method to supplement the position of the deleted outlier value; Normalizing the deviation of elements in the device state matrix by linear transformation; The principal component analysis method is used to perform dimension reduction processing on the equipment state matrix after the deviation is standardized to obtain the preprocessed data.

4. The method according to claim 1, characterized in that: The preprocessed data is processed by a closed-loop clustering algorithm to obtain data classification in different scenarios, including: Inputting the preprocessed data into a closed-loop clustering algorithm model to classify all scenes into several categories; Based on the measurement quantities of the scenes in the same category after clustering as the prediction objects, convolutional neural network models are trained respectively with the goal of minimizing the root mean square error, and prediction models corresponding to the categories after clustering are obtained; The data of each scene in all scenes are processed as follows: prediction is performed using the prediction models respectively to obtain multiple groups of prediction results equal to the number of prediction models; if a group of prediction results among the multiple groups of prediction results has the smallest root mean square error, the scene corresponding to the group of prediction results is moved to the corresponding category; Repeat the aforementioned steps of obtaining the prediction model and the data processing steps for each scenario until the termination condition is met, wherein the termination condition includes that the clustering result no longer changes or the maximum number of iterations is reached; The clustering results when the end condition is reached are used as the data classification in the different scenarios.

5. The method according to claim 4, characterized in that The method further comprises: Generate cluster labels based on the clustering results when the end condition is reached and add them to the preprocessed data; and The pre-processed data after data classification in different scenarios is unevenly processed.

6. The method according to claim 1, characterized in that The random forest algorithm in the combined model is configured as follows: the data in the training set is input into the random forest algorithm to obtain corresponding outputs, and the input data and output data of the random forest algorithm are correspondingly fused to obtain fused data; The convolutional neural network algorithm in the combined model is configured to: obtain a feature extraction result according to the input fusion data; The echo state network algorithm in the combined model is configured to generate a state matrix according to the feature extraction result, and the state matrix obtains a final output result through an output weight matrix.

7. The method according to claim 6, characterized in that The input data and the output data of the random forest algorithm are correspondingly fused to obtain fused data, including: the input data and the output data of the random forest algorithm are spliced ​​to obtain fused data; The echo state network algorithm is trained through the following steps: initializing the weight matrix and hyperparameters in the echo state network algorithm; generating corresponding feature extraction results after the data in the training set passes through the convolutional neural network part, and the feature extraction results are input into the echo state network part to train the connection weights from the hidden layer to the output layer; the hidden layer obtains the final output result through the output weight matrix of the output layer.

8. The method according to claim 1, characterized in that: The verified combined model is used to output the corresponding prediction results based on the input environmental state quantity and the corresponding measurement quantity, including: Obtain real-time data of environmental status and corresponding measurement quantities; constructing the real-time data into a state matrix form; The real-time data in the form of the state matrix is ​​input into the verified combined model, and the output of the verified combined model is used as the corresponding prediction result.

9. The method according to claim 1, characterized in that: After outputting a corresponding prediction result based on the input environmental state quantity and the corresponding metering quantity, the method further includes: Fitting the prediction results output by the verified combined model into a dynamic prediction curve; Calculating the residual between the dynamic prediction curve and the actual curve obtained by fitting the real-time data within a set time interval, and obtaining a residual dynamic change curve based on the residual; The equipment status of the substation is determined according to the fluctuation status of the residual dynamic change curve.

10. A device for predicting the state of a substation metering system, characterized in that: The device includes: A data acquisition module, used to acquire historical data of the substation including environmental status quantities and corresponding metering quantities; A preprocessing module, used for preprocessing the historical data to obtain preprocessed data; A data clustering module, used to process the preprocessed data through a closed-loop clustering algorithm to obtain data classification under different scenarios, and divide the preprocessed data after data classification into a training set and a validation set; and The combined model module is used to train the combined model using the training set and to verify the trained combined model using the verification set; the combined model includes a random forest algorithm, a convolutional neural network algorithm and an echo state network algorithm; the verified combined model is used to output a corresponding prediction result based on the input environmental state quantity and the corresponding metering quantity.

11. The device according to claim 10, characterized in that Obtain historical data of the substation including environmental status quantities and corresponding metering quantities, including: Collecting historical data of the substation, wherein the historical data includes environmental state quantities and corresponding metering quantities; The environmental state quantities and their corresponding measurement quantities are distinguished according to different scenarios, and the environmental state quantities and measurement quantities with corresponding relationships are combined into a state matrix form.

12. The device according to claim 10, characterized in that Preprocessing the historical data to obtain preprocessed data includes: constructing a device status matrix based on the historical data; Performing value interval judgment on the elements in the device state matrix, and determining the values ​​that are not in the preset value interval as abnormal values; After deleting the outlier value, a new value is generated by using the Lagrange interpolation method to supplement the position of the deleted outlier value; Normalizing the deviation of elements in the device state matrix by linear transformation; The principal component analysis method is used to perform dimension reduction processing on the equipment state matrix after the deviation is standardized to obtain the preprocessed data.

13. The device according to claim 10, characterized in that The preprocessed data is processed by a closed-loop clustering algorithm to obtain data classification in different scenarios, including: Inputting the preprocessed data into a closed-loop clustering algorithm model to classify all scenes into several categories; Based on the measurement quantities of the scenes in the same category after clustering as the prediction objects, convolutional neural network models are trained respectively with the goal of minimizing the root mean square error, and prediction models corresponding to the categories after clustering are obtained; The data of each scene in all scenes are processed as follows: prediction is performed using the prediction models respectively to obtain multiple groups of prediction results equal to the number of prediction models; if a group of prediction results among the multiple groups of prediction results has the smallest root mean square error, the scene corresponding to the group of prediction results is moved to the corresponding category; Repeat the aforementioned steps of obtaining the prediction model and the data processing steps for each scenario until the termination condition is met, wherein the termination condition includes that the clustering result no longer changes or the maximum number of iterations is reached; The clustering results when the end condition is reached are used as the data classification in the different scenarios.

14. The device according to claim 13, characterized in that The device further comprises an imbalance processing module, wherein the imbalance processing module is used for: Generate cluster labels based on the clustering results when the end condition is reached and add them to the preprocessed data; and The pre-processed data after data classification in different scenarios is unevenly processed.

15. The device according to claim 10, characterized in that The random forest algorithm in the combined model is configured as follows: the data in the training set is input into the random forest algorithm to obtain corresponding outputs, and the input data and output data of the random forest algorithm are correspondingly fused to obtain fused data; The convolutional neural network algorithm in the combined model is configured to: obtain a feature extraction result according to the input fusion data; The echo state network algorithm in the combined model is configured to generate a state matrix according to the feature extraction result, and the state matrix obtains a final output result through an output weight matrix.

16. The device according to claim 15, characterized in that The input data and the output data of the random forest algorithm are correspondingly fused to obtain fused data, including: the input data and the output data of the random forest algorithm are spliced ​​to obtain fused data; The echo state network algorithm is trained through the following steps: initializing the weight matrix and hyperparameters in the echo state network algorithm; generating corresponding feature extraction results after the data in the training set passes through the convolutional neural network part, and the feature extraction results are input into the echo state network part to train the connection weights from the hidden layer to the output layer; the hidden layer obtains the final output result through the output weight matrix of the output layer.

17. The device according to claim 10, characterized in that The verified combined model is used to output the corresponding prediction results based on the input environmental state quantity and the corresponding measurement quantity, including: Obtain real-time data of environmental status and corresponding measurement quantities; constructing the real-time data into a state matrix form; The real-time data in the form of the state matrix is ​​input into the verified combined model, and the output of the verified combined model is used as the corresponding prediction result.

18. The device according to claim 10, characterized in that The device further comprises a state determination module, wherein the state determination module is used to: after outputting a corresponding prediction result based on the input environmental state quantity and the corresponding metering quantity, Fitting the prediction results output by the verified combined model into a dynamic prediction curve; Calculating the residual between the dynamic prediction curve and the actual curve obtained by fitting the real-time data within a set time interval, and obtaining a residual dynamic change curve based on the residual; The equipment status of the substation is determined according to the fluctuation status of the residual dynamic change curve.

19. An electronic device, characterized in that: include: at least one processor; a memory connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the at least one processor implements the steps of the substation metering system state prediction method as described in any one of claims 1 to 9 by executing the instructions stored in the memory.

20. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the steps of the method for predicting the state of a substation metering system as claimed in any one of claims 1 to 9 are implemented.

21. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps of the method for predicting the state of a substation metering system as claimed in any one of claims 1 to 9 are implemented.

Images