Power failure classification method based on deep learning

Through a hybrid model of convolutional neural network and recurrent neural network, combined with Adam optimization algorithm, the problem of insufficient misjudgment, misjudgment and feature extraction capabilities in power failure classification is solved, and higher classification accuracy and real-timeness are achieved.

CN118916746APending Publication Date: 2024-11-08JIANGSU ELECTRIC POWER INFORMATION TECH
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202411091972.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-09
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The existing power fault classification technology has problems such as high misjudgment rate, high misjudgment rate, limited feature extraction capability, and insufficient real-time processing capability.

Method used

A hybrid model of convolutional neural network and recurrent neural network is adopted, combined with Adam optimization algorithm, data preprocessing and feature extraction are performed on power failure data, fault classification models are generated, and fault probability distribution is predicted through real-time data.

Benefits of technology

It improves the accuracy and real-time nature of fault classification, significantly improves the classification accuracy and efficiency, and effectively solves the shortcomings in the existing technology.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118916746A_ABST
    Figure CN118916746A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of power system fault diagnosis, and discloses a power fault classification method based on deep learning, which comprises the following steps: collecting fault data from a power system, and carrying out data preprocessing and feature extraction on the fault data to obtain target features; constructing a hybrid model based on a convolutional neural network and a recurrent neural network; optimizing the hybrid model by using an Adam optimization algorithm and the target features to generate a fault classification model; and inputting power system operation data collected in real time into the fault classification model for prediction about fault probability distribution, and outputting a fault classification result. According to the method, the hybrid model of the convolutional neural network and the recurrent neural network is combined, and the Adam optimization algorithm is used for optimization, so that the classification accuracy, the real-time performance and the robustness are greatly improved, the defects in fault classification in the prior art are effectively overcome, and the fault classification accuracy and efficiency are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power system fault diagnosis, and in particular to a power fault classification method based on deep learning. Background Art

[0002] As the scale and complexity of power systems continue to increase, fault diagnosis and classification technology has become the key to ensuring the stable operation of power systems. Traditional power fault diagnosis methods mainly rely on expert experience and rule bases. These methods often show high misjudgment and missed judgment rates when faced with complex and changeable fault conditions. In addition, with the advancement of data acquisition and storage technology in power systems, massive historical fault data and real-time operation data have made intelligent fault diagnosis possible.

[0003] In recent years, the rapid development of machine learning, especially deep learning technology, has provided new ideas for power fault classification. Convolutional neural network (CNN) performs well in image recognition and can effectively extract the spatial features of data; while recurrent neural network (RNN) has significant advantages in processing time series data; combining the two to build a hybrid model can make full use of the spatiotemporal characteristics of power fault data and improve the accuracy and robustness of fault classification.

[0004] However, the existing power fault classification technology still has some significant shortcomings. First, the traditional rule-based and expert system-based methods are unable to cope with new or complex faults and cannot be updated and adapted to the dynamically changing power system environment in a timely manner; second, although some methods based on shallow machine learning models can handle certain nonlinear relationships, their feature extraction capabilities are limited and cannot fully mine the deep information of the data; in addition, although some existing deep learning models have improved the accuracy of fault classification to a certain extent, there are still problems of misjudgment and missed judgment when dealing with multiple fault types and complex background noise.

[0005] Especially in real-time data processing and high-dimensional data feature extraction, the existing technology has not yet achieved ideal results. In contrast, the present invention optimizes the fault classification model by introducing a hybrid model of convolutional neural network and recurrent neural network and combining it with the Adam optimization algorithm. It can better adapt to complex and changeable power fault data and improve classification accuracy and efficiency. Summary of the invention

[0006] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract of the specification and the title of the invention of this application to avoid blurring the purpose of this section, the abstract of the specification and the title of the invention, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0007] In view of the above existing problems, the present invention is proposed to improve the accuracy and real-time performance of power fault classification.

[0008] Therefore, the technical problem solved by the present invention is that the existing power fault classification technology has the problems of high misjudgment rate, high missed judgment rate, limited feature extraction capability and insufficient real-time processing capability.

[0009] In order to solve the above technical problems, the present invention provides the following technical solutions: collecting fault data from the power system and performing data preprocessing and feature extraction on the fault data to obtain target features;

[0010] Build a hybrid model based on convolutional neural network and recurrent neural network;

[0011] Utilizing the Adam optimization algorithm and the target features, the hybrid model is optimized to generate a fault classification model;

[0012] The real-time collected power system operation data is input into the fault classification model to make a prediction on the fault probability distribution, and the fault classification result is output.

[0013] As a preferred solution of the power fault classification method based on deep learning described in the present invention, the collected fault data includes but is not limited to voltage data, current data, frequency data, power data, harmonic data, temperature data, and equipment status data;

[0014] By performing data cleaning, time stamp synchronization and data standardization on the fault data, a sample data set in a unified format is obtained.

[0015] As a preferred solution of the power fault classification method based on deep learning described in the present invention, the target feature is obtained, including:

[0016] Divide the preprocessed fault data into time windows, define the window size as 1 second to 10 seconds, and the step size as 0.5 second to 5 seconds;

[0017] Calculate the mean, standard deviation, kurtosis and skewness of the fault data in each window to obtain time domain features;

[0018] Perform Fourier transform on the fault data in each window, calculate its spectrum energy, main frequency and harmonic content, and obtain frequency domain features;

[0019] Performing wavelet transform on the fault data in each window, and extracting wavelet transform coefficients as time-frequency features;

[0020] Calculate the maximum value, minimum value, peak-to-peak value and root mean square value of the fault data in each window to obtain statistical characteristics;

[0021] The time domain features, the frequency domain features, the time-frequency features and the statistical features are combined into the target features.

[0022] As a preferred solution of the power fault classification method based on deep learning described in the present invention, a hybrid model is constructed based on a convolutional neural network and a recurrent neural network, and its mathematical expression formula is as follows:

[0023]

[0024] Among them, σ is the activation function, such as the Sigmoid function, n is the number of convolutional layers, a and b are the lower and upper limits of the integration interval, m is the number of convolution kernels, and w ij is the weight of the convolution kernel, φ is a nonlinear activation function, such as the ReLU function, p is the number of RNNs, ψ is the activation function of the RNN, such as the Tanh function, and x ijk is the input data, which represents the input of the i-th convolution layer, the j-th convolution kernel and the k-th RNN. is the Gaussian attenuation function used for filtering, ξ k is the weight of the kth RNN, c and d are the lower and upper limits of the integration interval, which are used for complex information filtering, and α is the adjustment parameter, which represents the weight of nonlinear adjustment.

[0025] As a preferred solution of the power fault classification method based on deep learning described in the present invention, generating the fault classification model includes:

[0026]

[0027] Among them, i is the index of the convolution layer, n is the number of convolution layers, j is the index of the convolution kernel, m is the number of convolution kernels, and w is the kernel index. ij is the weight of the convolution kernel, k is the index of the RNN, p is the number of RNNs, x ijk The input data represents the input of the jth convolution kernel of the i-th convolution layer and the k-th RNN, ξ k is the weight of the kth RNN, η ijk is the data after the convolution kernel, λ ij is the weight after the convolution layer, c and d are the lower and upper limits of the integral interval for complex information filtering, α is the adjustment parameter representing the weight of nonlinear adjustment, q is the step index of the Adam optimization algorithm, and γ l is the momentum of the Adam algorithm, β l is the scaling factor of the Adam algorithm.

[0028] As a preferred solution of the power fault classification method based on deep learning described in the present invention, prediction of fault probability distribution is performed, including:

[0029] The power system operation data collected in real time is input into the trained fault classification model after data preprocessing;

[0030] Perform forward propagation through each layer of the model, calculate the activation value of each layer, and finally get the result of the output layer;

[0031] The Softmax function is used in the output layer to process the results of the model output layer to obtain the probability distribution of each fault type. The mathematical expression formula is as follows:

[0032]

[0033] Where P(y=j|x) represents the probability that sample x belongs to fault category j, W j is the weight matrix, b is the bias, and K is the total number of categories;

[0034] According to the probability distribution, the fault type with the highest probability is selected as the final classification result.

[0035] As a preferred solution of the power fault classification method based on deep learning described in the present invention, the fault type with the highest probability is selected as the final classification result, and its mathematical expression formula is:

[0036]

[0037] in, The predicted fault type.

[0038] As a preferred solution of the deep learning-based power fault classification method described in the present invention, the fault types include but are not limited to line short circuit, equipment failure, overload, ground fault and harmonic distortion.

[0039] Beneficial effects of the invention: The invention achieves a significant improvement in classification accuracy, real-time performance and robustness by combining a hybrid model of a convolutional neural network and a recurrent neural network and optimizing using the Adam optimization algorithm, effectively solving the shortcomings of the prior art in fault classification and improving the accuracy and efficiency of fault classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0041] Figure 1 It is a flowchart of a power fault classification method based on deep learning shown in the present invention;

[0042] Figure 2 This is a schematic diagram of the fault probability distribution prediction results shown in the present invention. DETAILED DESCRIPTION

[0043] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0044] Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without making any creative work should fall within the scope of protection of the present invention.

[0045] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0046] According to an embodiment of the present invention, Figure 1 The flowchart shown is a method for classifying power faults based on deep learning, which specifically includes the following steps:

[0047] S1, collect fault data from the power system and perform data preprocessing and feature extraction to obtain target features;

[0048] S2, build a hybrid model based on convolutional neural network and recurrent neural network;

[0049] S3, using the Adam optimization algorithm and target features, optimize the hybrid model to generate a fault classification model;

[0050] S4. Input the real-time collected power system operation data into the fault classification model to predict the fault probability distribution and output the fault classification result.

[0051] Preferably, the present invention achieves a significant improvement in classification accuracy, real-time performance and robustness by combining a hybrid model of a convolutional neural network and a recurrent neural network and optimizing using the Adam optimization algorithm, effectively solving the deficiencies of the prior art in fault classification and improving the accuracy and efficiency of fault classification.

[0052] The following describes in more detail the implementation process and / or effects of certain embodiments of the present invention in conjunction with some preferred or optional examples of the present invention.

[0053] In an optional implementation manner, the collected fault data includes but is not limited to voltage data, current data, frequency data, power data, harmonic data, temperature data, and equipment status data; wherein:

[0054] The voltage data includes at least the phase voltage and line voltage, the current data includes at least the phase current and zero-sequence current, the frequency data is the operating frequency of the power grid, the power data includes at least the active power and reactive power, the harmonic data includes at least the current and voltage harmonic content, the temperature data includes at least the temperature of key equipment such as transformers and lines, and the equipment status data includes the switch status and protection action information.

[0055] By performing data cleaning, timestamp synchronization and data standardization on the fault data, a sample data set in a unified format is obtained.

[0056] As an example, standardization is performed so that its mean is 0 and its standard deviation is 1. The mathematical expression formula is as follows:

[0057]

[0058] Among them, z is the standardized data, x is the original data, μ is the data mean, and σ is the data standard deviation.

[0059] Furthermore, the target features are obtained, including:

[0060] The preprocessed fault data is divided into time windows, and the window size is defined as 1 second to 10 seconds, and the step size is 0.5 second to 5 seconds;

[0061] Calculate the mean, standard deviation, kurtosis, and skewness of the fault data in each window to obtain the time domain characteristics;

[0062] Perform Fourier transform on the fault data in each window, calculate its spectrum energy, main frequency and harmonic content, and obtain frequency domain characteristics;

[0063] Perform wavelet transform on the fault data in each window and extract the wavelet transform coefficients as time-frequency features;

[0064] Calculate the maximum value, minimum value, peak-to-peak value and RMS value of the fault data in each window to obtain statistical characteristics;

[0065] Time domain features, frequency domain features, time-frequency features and statistical features are combined into target features.

[0066] By preprocessing the collected data, noise and outliers can be effectively eliminated, and the most important features for fault classification, such as voltage, current, frequency and other key parameters, can be extracted to improve the quality of model input data and enhance the robustness and accuracy of the model.

[0067] It is not difficult to understand that data such as voltage, current, and frequency cover all aspects of the operating status of the power system. Through data cleaning, timestamp synchronization, and standardization processing, the consistency and accuracy of these data can be ensured, thereby providing high-quality data input for subsequent feature extraction and model training, and improving the quality of fault classification model input data and feature extraction effects.

[0068] Preferably, by dividing the preprocessed fault data into time windows and calculating features of multiple dimensions, multi-dimensional extraction and fusion of data features are achieved. By calculating time domain features (such as mean, standard deviation, etc.), frequency domain features (such as spectral energy, main frequency, etc.), time-frequency features (such as wavelet transform coefficients) and statistical features (such as maximum value, minimum value, etc.), the characteristics of the fault data are fully captured, thereby generating high-quality target features for model training, thereby improving the feature extraction accuracy and recognition effect of the fault classification model.

[0069] As an example, a hybrid model is constructed based on a convolutional neural network and a recurrent neural network. Its mathematical expression formula is as follows:

[0070]

[0071] Among them, σ is the activation function, such as the Sigmoid function, n is the number of convolutional layers, a and b are the lower and upper limits of the integration interval, m is the number of convolution kernels, and w ij is the weight of the convolution kernel, φ is a nonlinear activation function, such as the ReLU function, p is the number of RNNs, ψ is the activation function of the RNN, such as the Tanh function, and x ijk is the input data, which represents the input of the i-th convolution layer, the j-th convolution kernel and the k-th RNN. is the Gaussian attenuation function used for filtering, ξ k is the weight of the kth RNN, c and d are the lower and upper limits of the integration interval, which are used for complex information filtering, and α is the adjustment parameter, which represents the weight of nonlinear adjustment.

[0072] CNN is good at processing spatial data, such as edge and texture information in image data, while RNN is good at processing time series data, such as the timing patterns and trends of signals. By combining these two network structures, we can make full use of the spatial and temporal characteristics of power fault data, improve the model's ability to recognize complex fault modes, and improve the accuracy and generalization ability of the fault classification model.

[0073] As an example, a fault classification model is generated, including:

[0074]

[0075] Among them, i is the index of the convolution layer, n is the number of convolution layers, j is the index of the convolution kernel, m is the number of convolution kernels, and w is the kernel index.ij is the weight of the convolution kernel, k is the index of the RNN, p is the number of RNNs, x ijk The input data represents the input of the jth convolution kernel of the i-th convolution layer and the k-th RNN, ξ k is the weight of the kth RNN, η ijk is the data after the convolution kernel, λ ij is the weight after the convolution layer, c and d are the lower and upper limits of the integral interval for complex information filtering, α is the adjustment parameter representing the weight of nonlinear adjustment, q is the step index of the Adam optimization algorithm, and γ l is the momentum of the Adam algorithm, β l is the scaling factor of the Adam algorithm.

[0076] It should be noted that the Adam optimization algorithm combines the advantages of the momentum and RMSProp algorithms. It can adaptively adjust the learning rate and quickly converge to the optimal solution, thereby improving the training efficiency and effect of the model. By performing deep learning training on the target features of the fault data, an efficient fault classification model can be generated, which improves the model training efficiency and enhances the model performance and stability.

[0077] Specifically, the prediction of the failure probability distribution is carried out, including:

[0078] The real-time collected power system operation data is input into the trained fault classification model after data preprocessing;

[0079] Perform forward propagation through each layer of the model, calculate the activation value of each layer, and finally get the result of the output layer;

[0080] The Softmax function is used in the output layer to process the results of the model output layer to obtain the probability distribution of each fault type. The mathematical expression formula is as follows:

[0081]

[0082] Where P(y=j|x) represents the probability that sample x belongs to fault category j, W j is the weight matrix, b is the bias, and K is the total number of categories;

[0083] According to the probability distribution, the fault type with the highest probability is selected as the final classification result.

[0084] As an example, the fault type with the highest probability is selected as the final classification result, and its mathematical expression formula is:

[0085]

[0086] Output probability distribution and classification confidence:

[0087] Fault Probability Distribution={P(y=1|x), P(y=2|x),..., P(y=K|x)}

[0088]

[0089] in, The predicted fault type.

[0090] It should be noted that the confidence level of the classification result is provided as a basis for judging the reliability of the result, and the confidence level can directly take the maximum probability value.

[0091] By way of example, fault types include, but are not limited to, line short circuit, equipment failure, overload, ground fault, and harmonic distortion.

[0092] It should also be noted that in this embodiment, by integrating multiple deep learning technologies, an efficient fault classification model is generated to improve the accuracy and efficiency of fault classification. After forward propagation and Softmax function processing, the probability distribution of each fault type is obtained, and the fault type with the highest probability is selected as the final classification result, thereby improving the credibility of the fault classification result and the decision support capability.

[0093] In order to better illustrate the implementation and application of the above technical solution, the embodiments of the present invention are illustrated by the following practical application scenario examples:

[0094] (1) Data collection and preprocessing

[0095] The fault data collected in real time includes:

[0096] Voltage data: phase voltage and line voltage;

[0097] Current data: phase current and zero-sequence current;

[0098] Frequency data: grid operating frequency;

[0099] Power data: active power and reactive power;

[0100] Harmonic data: current and voltage harmonic content;

[0101] Temperature data: the temperature of key equipment such as transformers and lines;

[0102] Equipment status data: switch status, protection action information.

[0103] After data cleaning, time alignment and standardization, these data are divided into time windows (window size is 5 seconds and step size is 1 second) and input into the fault classification model.

[0104] (2) Fault probability distribution prediction

[0105] The preprocessed data is input into the trained fault classification model, and the probability distribution of each fault type is obtained through forward propagation and Softmax function;

[0106] After the input data of the model is processed, the following Figure 2 The probability distribution shown.

[0107] (3) Output fault classification results

[0108] Fault type: According to the probability distribution output by the Softmax function, the fault type with the highest probability is selected as the final classification result;

[0109] Predicted fault type: line short circuit;

[0110] Fault probability: Output the probability value corresponding to each fault type;

[0111] Probability distribution: line short circuit is 70%, equipment failure is 10%, overload is 5%, ground fault is 10%, harmonic distortion is 5%;

[0112] Classification confidence: Provides the confidence of the classification result as a basis for judging the reliability of the result. The confidence is the maximum probability value, such as classification confidence: 70%.

[0113] According to the fault type and probability distribution output by the model, combined with the expert system of the power system, fault diagnosis and processing suggestions are given; for example:

[0114] For line short circuit, it is recommended to check the line joints and insulation;

[0115] For equipment failure, it is recommended to repair or replace the corresponding equipment;

[0116] For overload, it is recommended to adjust the load distribution or increase the backup power supply.

[0117] Through the above steps and examples of actual application scenarios, it is illustrated that the real-time collected fault data is input into the fault classification model to predict the fault probability distribution, and output detailed fault classification results and corresponding processing suggestions to ensure the safe and stable operation of the power system.

[0118] The aforementioned data preprocessing, feature extraction and model training methods can be performed using methods and means in the prior art and will not be described in detail in this example.

[0119] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A power fault classification method based on deep learning, characterized in that: include: Collect fault data from the power system and perform data preprocessing and feature extraction to obtain target features; Build a hybrid model based on convolutional neural network and recurrent neural network; Utilizing the Adam optimization algorithm and the target features, the hybrid model is optimized to generate a fault classification model; The real-time collected power system operation data is input into the fault classification model to make a prediction on the fault probability distribution, and the fault classification result is output.

2. The power fault classification method based on deep learning according to claim 1 is characterized in that: The collected fault data includes but is not limited to voltage data, current data, frequency data, power data, harmonic data, temperature data, and equipment status data; By performing data cleaning, time stamp synchronization and data standardization on the fault data, a sample data set in a unified format is obtained.

3. The power fault classification method based on deep learning according to claim 1 is characterized in that: Obtaining the target feature includes: Divide the preprocessed fault data into time windows, define the window size as 1 second to 10 seconds, and the step size as 0.5 second to 5 seconds; Calculate the mean, standard deviation, kurtosis and skewness of the fault data in each window to obtain time domain features; Perform Fourier transform on the fault data in each window, calculate its spectrum energy, main frequency and harmonic content, and obtain frequency domain features; Performing wavelet transform on the fault data in each window, and extracting wavelet transform coefficients as time-frequency features; Calculate the maximum value, minimum value, peak-to-peak value and root mean square value of the fault data in each window to obtain statistical characteristics; The time domain features, the frequency domain features, the time-frequency features and the statistical features are combined into the target features.

4. The power fault classification method based on deep learning according to claim 1 is characterized in that: A hybrid model is constructed based on convolutional neural network and recurrent neural network. Its mathematical expression formula is as follows: Among them, σ is the activation function, such as the Sigmoid function, n is the number of convolutional layers, a and b are the lower and upper limits of the integration interval, m is the number of convolution kernels, and w ij is the weight of the convolution kernel, φ is a nonlinear activation function, such as the ReLU function, p is the number of RNNs, ψ is the activation function of the RNN, such as the Tanh function, and x ijk is the input data, which represents the input of the i-th convolution layer, the j-th convolution kernel and the k-th RNN. is the Gaussian attenuation function used for filtering, ξ k is the weight of the kth RNN, c and d are the lower and upper limits of the integration interval, which are used for complex information filtering, and α is the adjustment parameter, which represents the weight of nonlinear adjustment.

5. The power fault classification method based on deep learning according to claim 1 or 4, characterized in that: Generating the fault classification model includes: Among them, i is the index of the convolution layer, n is the number of convolution layers, j is the index of the convolution kernel, m is the number of convolution kernels, and w is the kernel index. ij is the weight of the convolution kernel, k is the index of the RNN, p is the number of RNNs, x ijk The input data represents the input of the jth convolution kernel of the i-th convolution layer and the k-th RNN, ξ k is the weight of the kth RNN, η ijk is the data after the convolution kernel, λ ij is the weight after the convolution layer, c and d are the lower and upper limits of the integral interval for complex information filtering, α is the adjustment parameter representing the weight of nonlinear adjustment, q is the step index of the Adam optimization algorithm, and γ l is the momentum of the Adam algorithm, β l is the scaling factor of the Adam algorithm.

6. The power fault classification method based on deep learning according to claim 1 is characterized in that: Make predictions about the probability distribution of failures, including: The power system operation data collected in real time is input into the trained fault classification model after data preprocessing; Perform forward propagation through each layer of the model, calculate the activation value of each layer, and finally get the result of the output layer; The Softmax function is used in the output layer to process the results of the model output layer to obtain the probability distribution of each fault type. The mathematical expression formula is as follows: Where P(y=j|x) represents the probability that sample x belongs to fault category j, W j is the weight matrix, b is the bias, and K is the total number of categories; According to the probability distribution, the fault type with the highest probability is selected as the final classification result.

7. The power fault classification method based on deep learning according to claim 6 is characterized in that: The fault type with the highest probability is selected as the final classification result, and its mathematical expression formula is: in, The predicted fault type.

8. The power fault classification method based on deep learning according to claim 7 is characterized in that: The fault types include but are not limited to line short circuit, equipment failure, overload, ground fault and harmonic distortion.

Citation Information

Cited By

  • Low-voltage distribution network operation state detection method based on fusion terminal

    CN119150210A

  • Defect identification method and system for petroleum accessory mud pump shell based on deep learning

    CN119198925A

  • EMB system-oriented high-frequency electric signal feature classification method and system

    CN121350852A