Method, device and equipment for determining partial discharge of power equipment and storage medium

Abstract

The application provides a kind of electric power equipment partial discharge determination method, device, equipment and storage medium, the target data obtained includes: the external partial discharge peak value of target electric power equipment, external partial discharge average, external partial discharge frequency, external partial discharge peak value difference, temperature, humidity, internal and external partial discharge peak value, internal and external partial discharge average, internal and external partial discharge frequency;Target data is input into multiple partial discharge prediction models to enable multiple partial discharge prediction models to process target data respectively, obtain multiple first determination results corresponding to the number of partial discharge prediction models, and then determine the final determination result;Wherein, the first determination result is used to indicate whether the target electric power equipment occurs partial discharge;Multiple partial discharge prediction models are obtained based on different learning algorithms training.The way can determine whether the target electric power equipment occurs partial discharge through multiple partial discharge prediction models, without relying on artificial experience, improve the accuracy and efficiency of judgment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system fault identification, and in particular to a method and device for determining partial discharge of power equipment, and a storage medium. BACKGROUND

[0002] Partial discharge (hereinafter referred to as partial discharge) is an important factor leading to insulation failure in power systems. It refers to the phenomenon of charge separation and recombination in a local area of electrical insulation material, which can cause partial damage and performance degradation of the insulation material. Therefore, monitoring and analyzing the partial discharge phenomenon is an important means to ensure the safe operation of power equipment. In practical applications, the frequency and intensity of partial discharge are affected by multiple factors, including temperature, humidity, current, etc.; by analyzing these influencing factors, the occurrence of partial discharge phenomenon can be better predicted, and appropriate processing measures can be taken in time. In related technologies, the partial discharge phenomenon is mainly analyzed and judged by artificial means, which relies more on artificial experience and reduces the accuracy and efficiency of judgment. SUMMARY

[0003] The purpose of the present application is to provide a method and device for determining partial discharge of power equipment, and a storage medium, to determine whether the target power equipment has partial discharge and improve the accuracy and efficiency of judgment.

[0004] The present application provides a method for determining partial discharge of power equipment, which comprises: obtaining target data of a target power equipment; wherein the target data includes: external partial discharge peak value, external partial discharge average value, external partial discharge frequency, external partial discharge peak value difference, temperature, humidity, internal and external partial discharge peak value, internal and external partial discharge average value, internal and external partial discharge frequency of the target power equipment; inputting the target data into a plurality of pre-trained partial discharge prediction models to process the target data, and obtaining a plurality of first determination results corresponding to the number of partial discharge prediction models; wherein the first determination result is used to indicate whether the target power equipment has partial discharge; the plurality of partial discharge prediction models are trained based on different learning algorithms; and determining a final determination result based on the plurality of first determination results.

[0005] Further, the step of determining the final determination result based on the plurality of first determination results comprises: weighting the plurality of first determination results according to a preset weighting coefficient to obtain the final determination result.

[0006] Further, each partial discharge prediction model is trained by: obtaining first sample data; pre-processing the first sample data to obtain second sample data; labeling a standard result for the second sample data to obtain labeled second sample data; wherein the standard result is used to represent: presence of partial discharge or absence of partial discharge; inputting the labeled second sample data into an initial first model, and using a target learning algorithm to predict a first result; wherein the target learning algorithm includes: a logistic regression algorithm, a decision tree algorithm, a random forest algorithm, a Gaussian naive Bayes algorithm, a linear discriminant analysis algorithm, a quadratic discriminant analysis algorithm, a support vector machine algorithm, and a k-nearest neighbor algorithm; calculating a loss value based on the first result and the standard result; training the first model based on the loss value to obtain the partial discharge prediction model.

[0007] Further, the standard result is determined by: obtaining internal and external partial discharge peak values corresponding to the second sample data; determining whether the preliminary classification result of the second sample data is suspicious partial discharge according to the internal and external partial discharge peak values corresponding to the second sample data; if the preliminary classification result of the second sample data is suspicious partial discharge, drawing a phase resolution partial discharge map using partial discharge phase data of the second sample data, and determining the standard result of the corresponding second sample data based on the phase resolution partial discharge map.

[0008] Further, the step of determining the standard result of the corresponding second sample data based on the phase resolution partial discharge map includes: obtaining a phase resolution partial discharge map corresponding to the second sample data; determining whether there are two amplitude peaks with a phase difference of 180° in the phase resolution partial discharge map; wherein each amplitude peak is greater than a preset amplitude threshold; if there are two amplitude peaks, determining that the standard result is: presence of partial discharge; if there are no two amplitude peaks, determining that the standard result is: absence of partial discharge.

[0009] Further, the step of training the first model based on the loss value to obtain the partial discharge prediction model includes: training the first model based on the loss value to obtain a trained first model; verifying the trained first model using a three-fold cross-validation method to obtain the partial discharge prediction model.

[0010] Further, the step of verifying the trained first model to obtain the partial discharge prediction model includes: obtaining a sample data set, and dividing the sample data set into three equal parts; wherein, in each part of the data set, the ratio of the number of sample data with partial discharge to the number of sample data without partial discharge is the same; selecting the first part of the data set as a test data set, and the other two parts of the data set as training data sets; training the initial first model using the training data sets to obtain a trained first model; using the trained first model to predict the test data set to obtain a first prediction result; selecting the second part of the data set as a test data set, and the other two parts of the data set as training data sets, and repeating the step of training the initial first model using the training data sets to obtain a second prediction result; selecting the third part of the data set as a test data set, and the other two parts of the data set as training data sets, and repeating the step of training the initial first model using the training data sets to obtain a third prediction result; and verifying the trained first model based on the first prediction result, the second prediction result, and the third prediction result to determine whether the trained first model is determined as the partial discharge prediction model.

[0011] The application provides a device for determining partial discharge of an electric power equipment, which comprises an acquisition module configured to acquire target data of a target electric power equipment, wherein the target data comprises external partial discharge peak value, external partial discharge average value, external partial discharge frequency, external partial discharge peak value difference, temperature, humidity, internal and external partial discharge peak value, internal and external partial discharge average value, and internal and external partial discharge frequency of the target electric power equipment; an output module configured to input the target data into a plurality of pre-trained partial discharge prediction models, so that the plurality of partial discharge prediction models process the target data respectively to obtain a plurality of first determination results corresponding to the number of partial discharge prediction models, wherein the first determination result is used to indicate whether the target electric power equipment has partial discharge; and a determination module configured to determine a final determination result based on the plurality of first determination results.

[0012] The application provides an electronic device, which comprises a processor and a memory, wherein the memory stores machine executable instructions capable of being executed by the processor, and the processor executes the machine executable instructions to implement the determination method of the partial discharge of the electric power equipment.

[0013] The application provides a machine readable storage medium, which stores machine executable instructions, wherein the machine executable instructions, when invoked and executed by a processor, cause the processor to implement the determination method of the partial discharge of the electric power equipment.

[0014] The application provides a method and device for determining partial discharge of an electric power equipment, and a storage medium. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the specific embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort.

[0016] Figure 1 A flow chart of a method for determining partial discharge of an electric power equipment is provided for the embodiments of the present application.

[0017] Figure 2 A flow chart of another method for determining partial discharge of an electric power equipment is provided for the embodiments of the present application.

[0018] Figure 3 A structural schematic diagram of a device for determining partial discharge of an electric power equipment is provided for the embodiments of the present application.

[0019] Figure 4 A structural schematic diagram of an electronic device is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0020] The technical solutions of the present application will be described in detail below with reference to the embodiments. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without any creative effort are within the scope of protection of the present application.

[0021] In the related art, the partial discharge phenomenon is mainly analyzed and judged by manual work, which is more dependent on manual experience, and reduces the accuracy and efficiency of judgment. Therefore, the embodiments of the present application provide a method, device and equipment for determining partial discharge of power equipment and a storage medium, which can be applied to the safe operation and early warning system of power equipment, especially in the scene of partial discharge detection and fault warning in power distribution stations.

[0022] In order to facilitate the understanding of the present embodiment, first, a method for determining partial discharge of power equipment is introduced, as shown in the following formula (I), which comprises the following steps: Figure 1

[0023] Step S102, obtaining target data of the target power equipment; wherein, the target data comprises: external partial discharge peak value, external partial discharge average value, external partial discharge frequency, external partial discharge peak value difference, temperature, humidity, internal and external partial discharge peak value, internal and external partial discharge average value, internal and external partial discharge frequency of the target power equipment;

[0024] The above-mentioned target power equipment can be a transformer, a distribution box, a switching device, etc. The above-mentioned external partial discharge peak value can be understood as the maximum detection value of the external partial discharge of the target power equipment within a preset time period. The above-mentioned external partial discharge average value can be understood as the average value of multiple detection values of the external partial discharge of the target power equipment within a preset time period. The above-mentioned external partial discharge frequency can be understood as the number of times of the external partial discharge of the target power equipment within a preset time period. The above-mentioned external partial discharge peak value difference can be understood as the difference between the maximum detection value and the minimum detection value of the external partial discharge of the target power equipment within a preset time period. The above-mentioned temperature can be understood as the temperature of the environment where the target power equipment is located. The above-mentioned humidity can be understood as the humidity of the environment where the target power equipment is located. The above-mentioned internal and external partial discharge peak value can be understood as the maximum detection value of the internal and external partial discharge of the target power equipment within a preset time period. The above-mentioned internal and external partial discharge average value can be understood as the average value of multiple detection values of the internal and external partial discharge of the target power equipment within a preset time period. The above-mentioned internal and external partial discharge frequency can be understood as the number of times of the internal and external partial discharge of the target power equipment within a preset time period. In actual implementation, when it is necessary to determine whether the target power equipment has partial discharge, the above-mentioned target data needs to be detected and obtained.

[0025] Step S104, inputting the target data into a plurality of pre-trained partial discharge prediction models, so that the plurality of partial discharge prediction models respectively process the target data, and obtaining a plurality of first determination results corresponding to the number of partial discharge prediction models; wherein, the first determination result is used to indicate whether the target power equipment has partial discharge; the plurality of partial discharge prediction models are trained based on different learning algorithms; ​

[0026] The partial discharge prediction model can be implemented by various convolutional neural networks, such as a residual network, a VGG network, etc.

[0027] The different learning algorithms can include a logistic regression algorithm, a decision tree algorithm, a random forest algorithm, a Gaussian naive Bayes algorithm, a linear discriminant analysis algorithm, a quadratic discriminant analysis algorithm, a support vector machine algorithm, a k-nearest neighbor algorithm, etc. A variety of different supervised learning algorithms can be selected according to actual needs. The logistic regression algorithm is an algorithm for binary classification problems. It outputs a probability value to determine the classification of samples by learning the relationship between the characteristics of the data and the target variable. It is suitable for processing linearly separable classification problems and estimates model parameters by maximizing the likelihood function. The decision tree algorithm represents the decision-making process of features by constructing a tree model, can handle continuous and discrete data, and is intuitive and easy to interpret. The decision tree model divides the data set into smaller subsets recursively until each subset (leaf node) contains only instances of the same class. Its advantage is simple and easy to understand, but it is prone to overfitting. The random forest algorithm is an algorithm that integrates multiple decision trees. It determines the final classification result by voting and has high accuracy and robustness. It improves the stability and accuracy of the model by constructing multiple decision trees and using the prediction results of each tree to vote. The Gaussian naive Bayes algorithm is based on Bayes' theorem and the assumption of feature independence, has high computational efficiency, and is suitable for processing large-scale data. It assumes that all features are independent of each other and classifies by calculating the conditional probability of each feature. The linear discriminant analysis algorithm is a supervised learning algorithm mainly used for data dimensionality reduction and classification. It finds the best projection direction to project high-dimensional data into low-dimensional space while maximizing the difference between classes and minimizing the distance between samples of the same class. The quadratic discriminant analysis algorithm is a discriminant analysis method based on Gaussian distribution. It treats sample data as observations of Gaussian distribution, estimates the mean vector and covariance matrix of each class, calculates the probability that the sample data belongs to each class, and discriminates according to the probability size. The support vector machine algorithm maximizes the interval between different classes by constructing a hyperplane and is suitable for high-dimensional data classification problems. The support vector machine algorithm finds an optimal partition hyperplane in high-dimensional space to maximize the classification interval, thereby improving the accuracy of classification. The k-nearest neighbor algorithm calculates the distance between the new sample and all samples in the training set, selects the k nearest samples for voting classification, and is a non-parametric method that does not require an explicit training process but has high computational complexity and is suitable for small-scale data sets.

[0028] In actual implementation, a plurality of different partial discharge prediction models can be trained by using a plurality of different learning algorithms. The target data obtained above can be respectively input into each partial discharge prediction model, and a corresponding first determination result can be respectively predicted by each partial discharge prediction model. The first determination results corresponding to different partial discharge prediction models can be the same or different. For example, the first determination result corresponding to a partial discharge prediction model trained by using a decision tree algorithm is normal, that is, the target power equipment does not have partial discharge; and the first determination result corresponding to a partial discharge prediction model trained by using a random forest algorithm is power failure, that is, the target power equipment has partial discharge.

[0029] In step S106, a final determination result is determined based on the plurality of first determination results.

[0030] In actual implementation, when the first determination results respectively predicted by the plurality of partial discharge prediction models are obtained, the final determination result can be obtained according to the plurality of first determination results. For example, a preset number threshold can be obtained in advance. When the number of determination results indicating that the target power equipment has partial discharge in the plurality of first determination results exceeds the preset number threshold, it can be considered that the final determination result is that the target power equipment has partial discharge. For another example, a weighting coefficient corresponding to each partial discharge prediction model can be allocated in advance according to actual requirements. The first determination result corresponding to the partial discharge prediction model with the highest weighting coefficient can be taken as the final determination result. Each first determination result can also be weighted and calculated according to the weighting coefficient to obtain the final determination result. The specific determination manner can be selected according to actual requirements, and is not limited herein.

[0031] The determination method of the partial discharge of the power equipment includes the following steps. Target data of a target power equipment is obtained. The target data includes an external partial discharge peak value, an external partial discharge average value, an external partial discharge frequency, an external partial discharge peak value difference, a temperature, a humidity, an internal and external partial discharge peak value, an internal and external partial discharge average value, and an internal and external partial discharge frequency of the target power equipment. The target data is input into a plurality of pre-trained partial discharge prediction models, so that the plurality of partial discharge prediction models process the target data to obtain a plurality of first determination results corresponding to the number of partial discharge prediction models. The first determination result is used to indicate whether the target power equipment has partial discharge. The plurality of partial discharge prediction models are trained based on different learning algorithms. A final determination result is determined based on the plurality of first determination results. The method can determine whether the target power equipment has partial discharge by using the plurality of partial discharge prediction models, without relying on artificial experience, and improves the accuracy and efficiency of judgment.

[0032] Another determination method of the partial discharge of the power equipment is also provided in the embodiments of the present application, which is implemented based on the above-mentioned embodiments.Figure 2 As shown, the method comprises the following steps:

[0033] Step S202, obtaining initial data of the target power equipment;

[0034] Step S204, preprocessing the initial data to obtain processed data;

[0035] In actual implementation, in order to improve the operation reliability of the target power equipment, the operation state of the target power equipment is usually monitored online, and the above-mentioned initial data usually includes various data obtained when the target power equipment is monitored, such as current flowing through insulation, environmental temperature, environmental humidity, partial discharge and other related data; The initial data monitored usually contains some invalid data or error data, such as if the environmental humidity is negative, it can be considered as invalid data, etc.; The detection standard corresponding to each detection item can be set in advance, and the data that does not meet the detection standard can be deleted to ensure the effectiveness of the processed data obtained.

[0036] Step S206, determining target data based on the processed data; wherein the target data includes: external partial discharge peak value, external partial discharge average value, external partial discharge frequency, external partial discharge peak value difference, temperature, humidity, internal and external partial discharge peak value, internal and external partial discharge average value, internal and external partial discharge frequency of the target power equipment;

[0037] The above-mentioned processed data may contain data irrelevant to judging whether partial discharge occurs, therefore, the above-mentioned target data can be extracted from the above-mentioned processed data, so as to judge whether the target power equipment occurs partial discharge according to these target data.

[0038] Step S208, inputting the target data into a plurality of pre-trained partial discharge prediction models, so that the plurality of partial discharge prediction models respectively process the target data to obtain a plurality of first determination results corresponding to the number of partial discharge prediction models; wherein the first determination result is used to indicate whether the target power equipment occurs partial discharge; the plurality of partial discharge prediction models are trained based on different learning algorithms;

[0039] Each partial discharge prediction model is trained by the following steps 40-45:

[0040] Step 40, obtaining first sample data;

[0041] Step 41, preprocessing the first sample data to obtain second sample data;

[0042] In actual implementation, in order to train each partial discharge prediction model, a training data set can be obtained in advance, which usually includes multiple pieces of sample data collected for multiple sample power equipment respectively. Each piece of sample data can be data collected in a preset time period, for example, 10 minutes of data is collected every 5 minutes. The first sample data can be 10 minutes of collected data. Similarly, there can be some invalid or erroneous data in the first sample data. The first sample data can be preprocessed to discard useless or erroneous information, clean up useless or unguaranteed information in the training data set, and ensure the data quality of the second sample data obtained. The second sample data usually includes: the external partial discharge peak value, the external partial discharge average value, the external partial discharge frequency, the external partial discharge peak value difference, the temperature, the humidity, the internal and external partial discharge peak value, the internal and external partial discharge average value, and the internal and external partial discharge frequency of the corresponding sample power equipment.

[0043] Step 42, labeling a standard result for the second sample data to obtain labeled second sample data; wherein the standard result is used to represent: existence of partial discharge or non-existence of partial discharge;

[0044] In actual implementation, after obtaining the second sample data, the corresponding standard result of the second sample data can be labeled according to the second sample data. The standard result can be used to represent whether the sample power equipment has partial discharge. For example, the standard result can be: normal, attention, early warning, and power off. Among them, “normal” indicates that there is no partial discharge, and “attention”, “early warning”, and “power off” all indicate that there is partial discharge, only the degree of partial discharge is different. Users can take corresponding measures according to the different degrees of partial discharge.

[0045] The standard result is determined through the following steps A-C:

[0046] Step A, obtaining the internal and external partial discharge peak value corresponding to the second sample data;

[0047] Step B, determining whether the preliminary classification result of the second sample data is suspicious partial discharge according to the internal and external partial discharge peak value corresponding to the second sample data;

[0048] Step C, if the preliminary classification result of the second sample data is suspicious partial discharge, drawing a phase resolution partial discharge diagram using the partial discharge phase data of the second sample data, and determining the standard result of the corresponding second sample data based on the phase resolution partial discharge diagram;

[0049] The above-mentioned phase-resolved partial discharge diagram (PRPD for short) is a widely used partial discharge analysis tool for describing the relationship between the power frequency phase of partial discharge occurrence, the discharge amount amplitude and the discharge times; in actual implementation, the second sample data can be preliminarily classified according to the preset judgment standard, specifically, the internal and external partial discharge peak values corresponding to the second sample data can be obtained, and it is judged whether the preliminary classification result of the second sample data is suspicious partial discharge by using the internal and external partial discharge peak values; the following standard can be referred to: when the internal and external partial discharge peak value corresponding to the second sample data is less than 8db, it is normal; when the internal and external partial discharge peak value corresponding to the second sample data is between 8db and 15db, it is concerned; when the internal and external partial discharge peak value corresponding to the second sample data is between 15db and 30db, it is pre-warning; when the internal and external partial discharge peak value corresponding to the second sample data is greater than 30db, it is power-off; in this process, 8db is temporarily determined as the boundary between partial discharge and non-partial discharge. For example, the internal and external partial discharge peak value with the highest detection value in the second sample data can be obtained, if the internal and external partial discharge peak value exceeds the pre-warning threshold, it is generally considered that the preliminary classification result of the second sample data is suspicious partial discharge, but since the accuracy of this determination method is poor, the data and its classification cannot be used as the data labeling of supervised learning only by the above-mentioned standard, for example, since the partial discharge peak value with high value may also be interference, the data and its classification cannot be used as the data labeling of supervised learning; in this case, further determination needs to be made by the phase-resolved partial discharge diagram to determine the standard result of the corresponding second sample data.

[0050] In specific implementation, the above-mentioned step of determining the standard result of the corresponding second sample data based on the phase-resolved partial discharge diagram can be implemented by the following steps one to three:

[0051] Step one, judging whether there are two amplitude peaks with a phase difference of 180° in the phase-resolved partial discharge diagram; wherein each amplitude peak is greater than a preset amplitude threshold;

[0052] The above-mentioned preset amplitude threshold can be set according to actual needs; the inventors have found through research that the phase-resolved partial discharge diagram generated by the sample power equipment with partial discharge phenomenon will have a relatively obvious feature, and the phase-resolved partial discharge diagram of surface partial discharge will generate two amplitude peaks at a phase distance of 180°, therefore, according to this judgment condition, whether the sample power equipment has partial discharge can be more accurately judged.

[0053] Step two, if there are two amplitude peaks, determining that the standard result is that there is partial discharge;

[0054] Step three, if there are no two amplitude peaks, determining that the standard result is that there is no partial discharge.

[0055] If there are two amplitude peaks with a phase difference of 180° in the phase resolution partial discharge diagram, it can be considered that the sample power equipment has partial discharge. If there are no two amplitude peaks with a phase difference of 180° in the phase resolution partial discharge diagram, it can be considered that the sample power equipment has no partial discharge. Since the preliminary classification result obtained above has poor accuracy, the preliminary classification result can be adjusted to obtain a standard result. For example, if the distribution box is disturbed, the partial discharge peak value may also exceed the 8db threshold set by humans and be mistakenly considered as having generated partial discharge. However, the phase resolution partial discharge diagram generated does not have two amplitude peaks with a phase difference of 180°, and therefore, it can be considered that no partial discharge is generated. After determining the standard result, simple manual confirmation or fine-tuning can be performed according to actual needs to obtain more accurate data labeling. In addition, when determining whether partial discharge occurs, the result can be determined not only by multiple models, but also by multiple partial discharge prediction models and image features of the phase resolution partial discharge diagram.

[0056] Step 43, input the labeled second sample data into the initial first model, and use a target learning algorithm to predict a first result; wherein the target learning algorithm includes: a logistic regression algorithm, a decision tree algorithm, a random forest algorithm, a Gaussian naive Bayes algorithm, a linear discriminant analysis algorithm, a quadratic discriminant analysis algorithm, a support vector machine algorithm, and a k-nearest neighbor algorithm.

[0057] The first model can be implemented by various convolutional neural networks, such as a residual network and a VGG network. In actual implementation, the labeled second sample data can be input into the initial first model, and different target learning algorithms can be used to output the first results corresponding to the labeled second sample data, respectively.

[0058] Step 44, calculate a loss value based on the first result and the standard result.

[0059] Step 45, train the first model based on the loss value to obtain a partial discharge prediction model.

[0060] The loss value can be used to evaluate the error between the first result and the standard result. In actual implementation, the first model can be trained based on the loss value between the first result and the standard result until the loss value converges, and a trained partial discharge prediction model is obtained.

[0061] Step 45 can be implemented by the following steps E and F:

[0062] Step E, train the first model based on the loss value to obtain a trained first model.

[0063] Step F, the trained first model is verified in a three-fold cross-validation manner to obtain the partial discharge prediction model.

[0064] The above step F can be implemented by the following steps four to step ten:

[0065] Step four, a sample data set is obtained, and the sample data set is evenly divided into three parts; wherein, in each data set, the ratio between the number of sample data with partial discharge and the number of sample data without partial discharge is the same;

[0066] Step five, the first data set is selected as the test data set, and the other two data sets are selected as the training data set;

[0067] The above three-fold cross-validation is one of the commonly used methods for evaluating the performance of machine learning models. The three-fold cross-validation method can effectively avoid the overfitting problem of the model on a specific data set, thereby more accurately evaluating the generalization ability of the model. In actual implementation, after training the first model according to the loss value between the above first result and the standard result to obtain the trained first model, in order to further improve the generalization ability of the final partial discharge prediction model, the sample data set can be divided into three parts as training data set and test data set respectively; specifically, in the embodiment, the sample data set can be evenly divided into three parts based on the three-fold cross-validation manner, which are data set A, data set B and data set C. In each data set, the ratio between the number of sample data with partial discharge and the number of sample data without partial discharge is the same, for example, each can account for half.

[0068] Step six, the initial first model is trained using the training data set to obtain the trained first model;

[0069] Step seven, the trained first model is used to predict the test data set to obtain the first prediction result;

[0070] Step eight, the second data set is selected as the test data set, and the other two data sets are selected as the training data set, and the step of training the initial first model using the training data set is repeated to obtain the second prediction result;

[0071] Step nine, the third data set is selected as the test data set, and the other two data sets are selected as the training data set, and the step of training the initial first model using the training data set is repeated to obtain the third prediction result;

[0072] Step ten, the trained first model is verified based on the first prediction result, the second prediction result and the third prediction result to determine whether the trained first model is determined as the partial discharge prediction model.

[0073] In a specific implementation, one of the data sets can be selected as a test data set, and the other two data sets can be used as training data sets to train the initial first model, and the process of training and testing is performed three times in total, and finally the performance of the obtained partial discharge prediction model is evaluated according to the average score of the three tests. Specifically, data set B and data set C can be used as training data sets to train the initial first model, the first prediction result can be obtained by using the trained first model to predict data set A, and the first test score of the trained first model can be evaluated according to the first prediction result; data set A and data set C can be used as training data sets to train the initial first model, the second prediction result can be obtained by using the trained first model to predict data set B, and the second test score of the trained first model can be evaluated according to the second prediction result; data set A and data set B can be used as training data sets to train the initial first model, the third prediction result can be obtained by using the trained first model to predict data set C, and the third test score of the trained first model can be evaluated according to the third prediction result; finally, the first test score, the second test score and the third test score can be averaged to verify the final performance measurement value of the trained first model, if the final performance measurement value meets the preset requirement, the finally trained first model can be used as the partial discharge prediction model, if the final performance measurement value does not meet the preset requirement, the above training process usually needs to be repeated until the final performance measurement value meets the preset requirement, and the finally trained first model is used as the partial discharge prediction model. The verification method can fairly evaluate the performance of the model, avoid the situation that a part of data is lucky or unlucky, cause the performance of the model to be overestimated or underestimated, and avoid the occurrence of special cases, and since the total data amount is very large, compared with more folds, three-fold cross-validation can obtain a more accurate model evaluation, reduce the calculation amount, and be more suitable for a big data scene.

[0074] In step S210, the plurality of first determination results are weighted and calculated according to a preset weighting coefficient to obtain a final determination result.

[0075] The preset weighting coefficients can be set according to actual needs. In actual implementation, a corresponding weighting coefficient can be set for each partial discharge prediction model. After obtaining the first determination result output by each partial discharge prediction model, the weighting coefficient can be used for weighting calculation to obtain a final determination result. If the first determination result output by the partial discharge prediction model is in the form of a category label, a corresponding numerical value can be set in advance for each category label corresponding to the partial discharge prediction model, so as to perform weighting calculation according to the numerical value corresponding to the first determination result. For example, the probability value output by the partial discharge prediction model obtained by using the logistic regression algorithm is 80%, indicating that the probability of occurrence of partial discharge is 80%. In order to facilitate weighting calculation, a corresponding numerical value can be set in advance for the category label that can be output by the partial discharge prediction model obtained by using the decision tree algorithm. For example, the category labels that can be output include “normal”, “warning” and “power failure”. The corresponding numerical value of the category label “normal” is 0, the corresponding numerical value of the category label “warning” is 0.5, and the corresponding numerical value of the category label “power failure” is 1. Therefore, in the weighting calculation, the corresponding numerical value 0.5 of the category label “warning” can be used for calculation.

[0076] For the convenience of understanding, the application of each algorithm in the embodiment is further described as follows:

[0077] 1. Logistic Regression

[0078] (1) Application purpose

[0079] The logistic regression is used for binary classification problem, that is, judging whether there is partial discharge phenomenon in the power system. The logistic regression learns the relationship between input features and partial discharge, and outputs the probability of occurrence of partial discharge.

[0080] (2) Input value

[0081] External partial discharge peak value, external partial discharge average value, external partial discharge frequency, external partial discharge peak value difference, temperature, humidity, internal and external partial discharge peak value, internal and external partial discharge average value, internal and external partial discharge frequency.

[0082] (3) Output value

[0083] The output value is a probability value, which indicates whether a certain instance (corresponding to the target data) is a partial discharge phenomenon. Based on the probability value, it can be determined whether to perform warning or take measures.

[0084] 2. Decision Tree

[0085] (1) Application purpose

[0086] For classifying partial discharge events, automatically determining the state of the power system (e.g., "normal", "attention", "warning" or "power outage"). Decision tree algorithm helps make decisions through conditional judgments, suitable for handling complex nonlinear relationships.

[0087] (2) Input values

[0088] External partial discharge peak value, external partial discharge average value, external partial discharge frequency, external partial discharge peak value difference, temperature, humidity, internal and external partial discharge peak value, internal and external partial discharge average value, internal and external partial discharge frequency.

[0089] (3) Output values

[0090] The output value is a category label (such as "normal", "warning" or "power outage"), which is decided according to the tree structure and indicates the running state of the device.

[0091] 3. Random Forest algorithm (Random Forest)

[0092] (1) Application purpose

[0093] As an ensemble method of decision trees, Random Forest improves the accuracy and robustness of prediction through the voting results of multiple decision trees. It helps to obtain more stable classification results in various partial discharge scenarios and is suitable for handling complex partial discharge detection tasks.

[0094] (2) Input values

[0095] External partial discharge peak value, external partial discharge average value, external partial discharge frequency, external partial discharge peak value difference, temperature, humidity, internal and external partial discharge peak value, internal and external partial discharge average value, internal and external partial discharge frequency.

[0096] (3) Output values

[0097] The output value is a category label (such as "normal", "attention", "warning" or "power outage"), which is decided according to the tree structure and indicates the running state of the device.

[0098] 4. Support Vector Machine algorithm (Support Vector Machine, abbreviated as SVM)

[0099] (1) Application purpose

[0100] Used to find the best hyperplane in high-dimensional space to maximize the separation between different classes (such as "partial discharge" and "non-partial discharge"). SVM is suitable for scenarios with complex feature space and small class distance, and can efficiently classify partial discharge.

[0101] (2) Input values

[0102] External PD peak value, external PD mean value, external PD frequency, external PD peak value difference, temperature, humidity, internal and external PD peak value, internal and external PD mean value, internal and external PD frequency.

[0103] (3) Output values

[0104] The output value is a category label (e.g., "partial discharge" or "non-partial discharge"), and by maximizing the separation between categories, it determines whether there is partial discharge in the power system.

[0105] 5. Gaussian Naive Bayes

[0106] (1) Application purpose

[0107] Based on Bayes' theorem, it uses the conditional probability of each feature to predict whether partial discharge occurs in the power system. Gaussian Naive Bayes is suitable for cases where features are relatively independent and the data volume is large.

[0108] (2) Input values

[0109] External PD peak value, external PD mean value, external PD frequency, external PD peak value difference, temperature, humidity, internal and external PD peak value, internal and external PD mean value, internal and external PD frequency.

[0110] (3) Output values

[0111] The output value is a category label (e.g., "partial discharge" or "non-partial discharge"), and by calculating the posterior probability of each category, it makes predictions.

[0112] 6. Linear Discriminant Analysis (LDA)

[0113] (1) Application purpose

[0114] LDA finds an optimal linear hyperplane to separate samples of different categories, suitable for cases where there is a linear relationship between features. It can help determine whether there is partial discharge in the power system and classify it.

[0115] (2) Input values

[0116] External PD peak value, external PD mean value, external PD frequency, external PD peak value difference, temperature, humidity, internal and external PD peak value, internal and external PD mean value, internal and external PD frequency.

[0117] (3) Output values

[0118] The output value is a class label (e.g., "normal" or "partial discharge"), and the classification is determined by maximizing the difference between classes.

[0119] 7、Quadratic Discriminant Analysis (QDA)

[0120] (1) Application purpose

[0121] QDA is an extension of LDA, suitable for more complex non-linear relationships between features. It realizes classification by fitting different Gaussian distributions for each class, and can more flexibly handle the differences in feature distribution of different classes.

[0122] (2) Input value

[0123] External partial discharge peak value, external partial discharge average value, external partial discharge frequency, external partial discharge peak value difference, temperature, humidity, internal and external partial discharge peak value, internal and external partial discharge average value, internal and external partial discharge frequency.

[0124] (3) Output value

[0125] The output value is a class label (e.g., "partial discharge" or "non-partial discharge"), and the classification is determined by discriminating the Gaussian distribution of different classes to determine whether the power equipment has partial discharge.

[0126] 8、K-Nearest Neighbors (KNN)

[0127] (1) Application purpose

[0128] KNN makes classification judgments based on the K nearest neighbors by calculating the distance between the new sample and all samples in the training data. It is a non-parametric method that does not require assumptions about the distribution of data, and is suitable for complex partial discharge pattern recognition in power systems.

[0129] (2) Input value

[0130] External partial discharge peak value, external partial discharge average value, external partial discharge frequency, external partial discharge peak value difference, temperature, humidity, internal and external partial discharge peak value, internal and external partial discharge average value, internal and external partial discharge frequency.

[0131] (3) Output value

[0132] The output value is a class label (e.g., "partial discharge" or "non-partial discharge"), and the classification is determined by distance measurement, based on the voting results of K neighbors to determine which class the sample belongs to.

[0133] A determination result table of partial discharge of a power equipment as shown in Table 1, wherein five groups of data are included, each group of data includes: external partial discharge peak value, external partial discharge average value, external partial discharge frequency, external partial discharge peak value difference, temperature, humidity, internal and external partial discharge peak value, internal and external partial discharge average value, internal and external partial discharge frequency, after each group of data is input into a plurality of partial discharge prediction models pre-trained, the final determination result corresponding to each group of data can be obtained, that is, "False" in Table 1, indicating that partial discharge occurs.

[0134] Table 1 is a determination result table of partial discharge of a power equipment

[0135]

[0136] The method has a judgment accuracy of 98.5%-99.3% through three-fold cross-validation test by using a logistic regression algorithm, a decision tree algorithm, a random forest algorithm, a support vector machine algorithm, a Gaussian naive Bayes algorithm and a k-nearest neighbor algorithm. The method can generate a partial discharge analysis report by analyzing the prediction results of the model. The partial discharge prediction models used above all use "external partial discharge peak value", "external partial discharge average value", "external partial discharge frequency", "external partial discharge peak value difference", "temperature", "humidity", "internal and external partial discharge peak value", "internal and external partial discharge average value", "internal and external partial discharge frequency" as input values, and predict whether partial discharge occurs. The values used here are all monitoring values that can be obtained in time, and can be judged in real time by using a plurality of partial discharge prediction models, so as to realize automatic early warning and assist field workers. Among them, the decision tree algorithm has an average accuracy of 99.1%; the Gaussian naive Bayes algorithm has an average accuracy of 98.5%; the random forest algorithm has an average accuracy of 99.3%; the support vector machine algorithm has an average accuracy of 99.1%; and the k-nearest neighbor algorithm has an average accuracy of 99.3%. The method can be applied to power system monitoring and fault diagnosis, and can monitor, analyze and predict the partial discharge phenomenon in the power system by combining data analysis and machine learning methods, and has high accuracy and efficiency. The method realizes accurate early warning and processing of key problems, ensures the safe operation of power equipment, and meets the realization of AI (Artificial Intelligence, Artificial Intelligence) agent partial discharge early warning of a power distribution station.

[0137] The embodiment of the application provides a determination device for partial discharge of a power equipment, as shown in Figure 3As shown, the device comprises: an acquisition module 30, configured to acquire target data of a target power equipment; wherein the target data comprises: an external partial discharge peak value, an external partial discharge average value, an external partial discharge frequency, an external partial discharge peak value difference, a temperature, a humidity, an internal and external partial discharge peak value, an internal and external partial discharge average value, and an internal and external partial discharge frequency of the target power equipment; an output module 31, configured to input the target data into a plurality of pre-trained partial discharge prediction models, so that the plurality of partial discharge prediction models respectively process the target data to obtain a plurality of first determination results corresponding to the number of partial discharge prediction models; wherein the first determination result is used to indicate whether the target power equipment has partial discharge; the plurality of partial discharge prediction models are trained based on different learning algorithms; and a determination module 32, configured to determine a final determination result based on each first determination result.

[0138] The above-mentioned device for determining partial discharge of power equipment can determine whether the target power equipment has partial discharge through the plurality of partial discharge prediction models, without relying on artificial experience, thereby improving the accuracy and efficiency of judgment.

[0139] Further, the determination module 32 is further configured to: perform weighted calculation on the plurality of first determination results according to a preset weighting coefficient to obtain the final determination result.

[0140] Further, the device further comprises: a model training module, each partial discharge prediction model is trained by the model training module, and the model training module is configured to: acquire first sample data; pre-process the first sample data to obtain second sample data; label a standard result of the second sample data to obtain labeled second sample data; wherein the standard result is used to represent: existence of partial discharge or non-existence of partial discharge; input the labeled second sample data into an initial first model, and predict a first result by using a target learning algorithm; wherein the target learning algorithm comprises: a logistic regression algorithm, a decision tree algorithm, a random forest algorithm, a Gaussian naive Bayes algorithm, a linear discriminant analysis algorithm, a quadratic discriminant analysis algorithm, a support vector machine algorithm, and a k-nearest neighbor algorithm; calculate a loss value based on the first result and the standard result; and train the first model based on the loss value to obtain the partial discharge prediction model.

[0141] Further, the device further comprises a standard result determination module, and the standard result determination module is configured to: acquire an internal and external partial discharge peak value corresponding to the second sample data; determine whether a preliminary classification result of the second sample data is suspicious partial discharge according to the internal and external partial discharge peak value corresponding to the second sample data; if the preliminary classification result of the second sample data is suspicious partial discharge, draw a phase-resolved partial discharge map by using partial discharge phase data of the second sample data, and determine a standard result of the corresponding second sample data based on the phase-resolved partial discharge map.

[0142] Further, the standard result determination module is further configured to: determine whether two amplitude peak values with a phase difference of 180° exist in the phase resolution partial discharge diagram; wherein each amplitude peak value is greater than a preset amplitude threshold value; if the two amplitude peak values exist, determine that the standard result is that the partial discharge exists; and if the two amplitude peak values do not exist, determine that the standard result is that the partial discharge does not exist.

[0143] Further, the model training module is further configured to: train the first model based on the loss value to obtain a trained first model; and verify the trained first model in a three-fold cross-validation manner to obtain the partial discharge prediction model.

[0144] Further, the model training module is further configured to: obtain a sample data set, and divide the sample data set into three equal parts; wherein in each part of the data set, the ratio between the number of sample data with partial discharge and the number of sample data without partial discharge is the same; select a first part of the data set as a test data set, and select the other two parts of the data set as training data sets; train the initial first model using the training data sets to obtain a trained first model; use the trained first model to predict the test data set to obtain a first prediction result; select a second part of the data set as a test data set, and select the other two parts of the data set as training data sets, and repeat the step of training the initial first model using the training data sets to obtain a second prediction result; select a third part of the data set as a test data set, and select the other two parts of the data set as training data sets, and repeat the step of training the initial first model using the training data sets to obtain a third prediction result; and verify the trained first model based on the first prediction result, the second prediction result, and the third prediction result to determine whether the trained first model is determined as the partial discharge prediction model.

[0145] The determination device for partial discharge of the power equipment provided in the embodiment of the present application has the same implementation principle and technical effects as the foregoing determination method for partial discharge of the power equipment, and for brevity of description, the part of the determination device for partial discharge of the power equipment that is not mentioned can refer to the corresponding content in the foregoing determination method for partial discharge of the power equipment.

[0146] The embodiment of the present application further provides an electronic device, referring to Figure 4 As shown in the figure, the electronic device includes a processor 130 and a memory 131, the memory 131 stores machine executable instructions that can be executed by the processor 130, and the processor 130 executes the machine executable instructions to implement the foregoing determination method for partial discharge of the power equipment.

[0147] Further, Figure 4The electronic device also includes a bus 132 and a communication interface 133, the processor 130, the communication interface 133 and the memory 131 are connected through the bus 132.

[0148] The memory 131 can include a high-speed random access memory (RAM), and can also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 133 (which can be wired or wireless), and the Internet, a wide area network, a local area network, a metropolitan area network, etc. can be used. The bus 132 can be an ISA bus, a PCI bus, or an EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one bidirectional arrow is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0149] The processor 130 can be an integrated circuit chip with signal processing capability. In the implementation process, each step of the above method can be completed by integrated logic circuits of hardware in the processor 130 or instructions in the form of software. The above processor 130 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. Each method, step and logic block disclosed in the embodiment of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in combination with the embodiment of the present application can be directly embodied as a hardware code processor for execution, or a combination of hardware and software modules in the code processor for execution. The software module can be located in a random access memory, a flash memory, a read-only memory, a programmable read-only memory, an electrically erasable programmable memory, a register, etc. The storage medium is mature in the art. The storage medium is located in the memory 131, and the processor 130 reads the information in the memory 131, and combines the hardware to complete the steps of the method of the above embodiment.

[0150] The embodiment of the present application further provides a machine readable storage medium which stores machine executable instructions, when the machine executable instructions are invoked and executed by a processor, the machine executable instructions cause the processor to implement the method for determining partial discharge of power equipment, and the specific implementation can be referred to the method embodiment, and will not be repeated here.

[0151] The method, device, equipment and storage medium of the present application can be used to determine partial discharge of power equipment, and the specific implementation can be referred to the method embodiment, and will not be repeated here.

[0152] The function can be stored in a computer readable storage medium if the function is realized in the form of a software function unit and sold or used as an independent product. Based on this understanding, the technical solutions of the present application or the part of the technical solutions that make essential contributions to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0153] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method of determining partial discharge of a power apparatus, characterized by, The method comprises: acquiring target data of a target power device; wherein the target data comprises: an external partial discharge peak value, an external partial discharge average value, an external partial discharge frequency, an external partial discharge peak value difference, a temperature, a humidity, an internal and external partial discharge peak value, an internal and external partial discharge average value, and an internal and external partial discharge frequency of the target power device; wherein the internal and external partial discharge peak value is the maximum detection value of internal and external partial discharges of the target power device within a preset time period; the internal and external partial discharge average value is the average value of multiple detection values of internal and external partial discharges of the target power device within the preset time period; and the internal and external partial discharge frequency is the number of times of internal and external partial discharges of the target power device within the preset time period; inputting the target data into a plurality of pre-trained partial discharge prediction models, so that the plurality of partial discharge prediction models respectively process the target data, and a plurality of first determination results corresponding to the number of partial discharge prediction models are acquired; wherein the first determination result is used to indicate whether the target power device has a partial discharge; and the plurality of partial discharge prediction models are trained based on different learning algorithms; determining a final determination result based on the plurality of first determination results; each partial discharge prediction model is trained in the following manner: acquiring first sample data; preprocessing the first sample data to obtain second sample data; labeling a standard result for the second sample data to obtain labeled second sample data; wherein the standard result is used to represent: there is a partial discharge or there is no partial discharge; inputting the labeled second sample data into an initial first model, and using a target learning algorithm to predict a first result; wherein the target learning algorithm comprises: a logistic regression algorithm, a decision tree algorithm, a random forest algorithm, a Gaussian naive Bayes algorithm, a linear discriminant analysis algorithm, a quadratic discriminant analysis algorithm, a support vector machine algorithm, and a k-nearest neighbor algorithm; calculating a loss value based on the first result and the standard result; training the first model based on the loss value to obtain the partial discharge prediction model; wherein the standard result is determined based on an internal and external partial discharge peak value and a phase resolution partial discharge map corresponding to the second sample data.

2. The method of claim 1, wherein, The step of determining a final determination result based on the plurality of first determination results comprises: performing weighted calculation on the plurality of first determination results according to a preset weighting coefficient to obtain the final determination result.

3. The method of claim 1, wherein, The standard result is determined in the following manner: acquiring an internal and external partial discharge peak value corresponding to the second sample data; determining whether a preliminary classification result of the second sample data is suspicious partial discharge according to the internal and external partial discharge peak value corresponding to the second sample data; if the preliminary classification result of the second sample data is suspicious partial discharge, drawing a phase resolution partial discharge map using partial discharge phase data of the second sample data, and determining a standard result of the corresponding second sample data based on the phase resolution partial discharge map.

4. The method of claim 3, wherein, The step of determining a standard result of the corresponding second sample data based on the phase resolution partial discharge map comprises: determining whether two amplitude peaks with a phase difference of 180° exist in the phase resolution partial discharge diagram; wherein each of the amplitude peaks is greater than a preset amplitude threshold value; if the two amplitude peaks exist, determining that the standard result is that partial discharge exists; if the two amplitude peaks do not exist, determining that the standard result is that partial discharge does not exist.

5. The method of claim 1, wherein, The step of training the first model based on the loss value to obtain the partial discharge prediction model comprises: training the first model based on the loss value to obtain a trained first model; verifying the trained first model in a three-fold cross-validation manner to obtain the partial discharge prediction model.

6. The method of claim 5, wherein, The step of training the first model based on the loss value to obtain the partial discharge prediction model comprises: obtaining a sample data set, and dividing the sample data set into three equal parts; wherein in each data set, the ratio between the number of sample data with partial discharge and the number of sample data without partial discharge is the same; selecting a first data set as a test data set, and the other two data sets as training data sets; training an initial first model using the training data sets to obtain a trained first model; using the trained first model to predict the test data set to obtain a first prediction result; selecting a second data set as a test data set, and the other two data sets as training data sets, and repeating the step of training the initial first model using the training data sets to obtain a second prediction result; selecting a third data set as a test data set, and the other two data sets as training data sets, and repeating the step of training the initial first model using the training data sets to obtain a third prediction result; verifying the trained first model based on the first prediction result, the second prediction result and the third prediction result to determine whether to determine the trained first model as the partial discharge prediction model.

7. An apparatus for determining partial discharge of a power device, characterized by The device comprises: an acquisition module configured to acquire target data of a target power equipment; wherein the target data comprises external partial discharge peak value, external partial discharge average value, external partial discharge frequency, external partial discharge peak value difference, temperature, humidity, internal and external partial discharge peak value, internal and external partial discharge average value, and internal and external partial discharge frequency of the target power equipment; wherein the internal and external partial discharge peak value is the maximum detection value of internal and external partial discharge of the target power equipment within a preset time period; the internal and external partial discharge average value is the average value of multiple detection values of internal and external partial discharge of the target power equipment within the preset time period; and the internal and external partial discharge frequency is the number of times of internal and external partial discharge of the target power equipment within the preset time period. An output module is configured to input the target data into a plurality of pre-trained partial discharge prediction models, so that the plurality of partial discharge prediction models respectively process the target data to obtain a plurality of first determination results corresponding to the number of partial discharge prediction models; wherein the first determination result is used to indicate whether the target power equipment has partial discharge; and the plurality of partial discharge prediction models are trained based on different learning algorithms; A determination module is configured to determine a final determination result based on the plurality of first determination results; A model training module is configured to obtain first sample data; The first sample data is preprocessed to obtain second sample data; The second sample data is labeled with a standard result to obtain labeled second sample data; wherein the standard result is used to represent that there is partial discharge or there is no partial discharge; The labeled second sample data is input into an initial first model, and a target learning algorithm is used to predict a first result; wherein the target learning algorithm includes a logistic regression algorithm, a decision tree algorithm, a random forest algorithm, a Gaussian naive Bayes algorithm, a linear discriminant analysis algorithm, a quadratic discriminant analysis algorithm, a support vector machine algorithm, and a k-nearest neighbor algorithm; A loss value is calculated based on the first result and the standard result; The first model is trained based on the loss value to obtain the partial discharge prediction model; wherein the standard result is determined based on internal and external partial discharge peak values and phase resolution partial discharge maps corresponding to the second sample data.

8. An electronic device, comprising: A processor and a memory are included, the memory stores machine executable instructions that can be executed by the processor, and the processor executes the machine executable instructions to implement the determination method of the power equipment partial discharge according to any one of claims 1-6.

9. A machine-readable storage medium, characterized in that, The machine readable storage medium stores machine executable instructions, and when the machine executable instructions are called and executed by the processor, the machine executable instructions cause the processor to implement the determination method of the power equipment partial discharge according to any one of claims 1-6.

Images