DSCNN-based in-oil arc discharge mode lightweight identification method

Through the lightweight deep-separable convolutional neural network model based on DSCNN, the time-frequency graph analysis of arc discharge signals is solved, and the problem of slow recognition speed of short gap arc discharge patterns in the existing technology is realized, and a fast, light and simple recognition method is realized. Suitable for resource-constrained environments.

CN119044690AActive Publication Date: 2024-11-29XI AN JIAOTONG UNIV

Patent Information

Application Number
CN202411106251.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-13
Publication Date
2024-11-29
Estimated Expiration
2044-08-13

AI Technical Summary

Technical Problem

In the prior art, the short gap arc discharge pattern recognition method for arc discharge faults is slow to detect and has a large calculation scale, so it cannot effectively cope with the challenge of arc discharge faults, especially in embedded systems or mobile maintenance equipment with limited computing resources, which is difficult to Achieve rapid and accurate positioning of defective parts and optimize maintenance strategies.

Method used

The time-frequency map of arc discharge signals is mined by lightweight deep-separable convolutional neural network model based on DSCNN. The time-frequency map is generated through wavelet transformation, and the model is trained through training sets and test sets, using regularization processing and improved activation. Function, build a lightweight recognition method to achieve fast, light and simple discharge pattern recognition.

Benefits of technology

It realizes that while ensuring high recognition accuracy, the inference speed is improved, the requirements for equipment computing capabilities are reduced, and the rapid demand for online monitoring for arc discharge pattern recognition is met. It is suitable for resource-constrained environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119044690A_ABST
    Figure CN119044690A_ABST
Patent Text Reader

Abstract

The invention discloses a DSCNN-based in-oil arc discharge mode lightweight identification method, and the method comprises the steps: carrying out an oil-immersed power transformer short-gap in-oil arc discharge test experiment; the method comprises the following steps: collecting discharge time domain data of an arc discharge test experiment by using a sensor, and performing wavelet transform on the discharge time domain data to generate a time-frequency map; dividing the time-frequency atlas into a training set and a test set, training a lightweight depth separable convolutional neural network model by using the training set, detecting the trained model by using the test set, and recording a first training parameter; using the training set to train a conventional convolutional neural network model, using the test set to detect the trained model, and recording a second training parameter; by comparing the first training parameter with the second training parameter, the DSCNN-based short-gap arc discharge mode recognition method is proved to have the advantage of light weight, and the requirement of online monitoring is met.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of arc discharge of oil-immersed power transformers, and in particular to a lightweight identification method for arc discharge patterns in oil based on DSCNN. Background Art

[0002] Arc discharge is one of the most serious faults in oil-immersed power transformers, posing a significant threat to national economic development and power system safety. By the time arc discharge reaches the long-gap arc discharge stage, the accumulated arc discharge is already substantial, making serious equipment failure almost inevitable. To reduce safety hazards and equipment damage, prevention measures must be taken before the discharge reaches the short-gap arc discharge or even spark discharge stage. When an arc discharge fault occurs in a transformer, the violent reaction generates various physical and chemical signals, including acoustic, optical, thermal, and other signals. Using online monitoring technology to acquire these signals in real time and explore their generation mechanisms and evolution patterns is a key approach to preventive maintenance.

[0003] Furthermore, discharge pattern recognition can quickly and accurately locate defective areas and optimize maintenance strategies in a timely manner. In the field of pattern recognition, artificial intelligence algorithms based on machine learning / deep learning have been widely studied, such as k-means clustering, support vector machines, Markov models, k-nearest neighbors, and random decision forests. Given the high risk of arc discharge, rapid identification of its fault type is crucial. However, in certain application scenarios, such as embedded systems or mobile maintenance equipment, available computing resources and storage space are very limited. This requires arc discharge fault diagnosis models to ensure high accuracy while also focusing on reducing model complexity and data size to enable operation within limited resources.

[0004] At present, research on arc discharge diagnosis mainly focuses on long-gap arc discharge, and there is a lack of research on pattern recognition of short-gap arc discharge and arc discharge caused by different defects. In the existing technology, discharge pattern recognition methods are mostly targeted at the field of partial discharge, with slow detection speed, large calculation scale, and high requirements on computer configuration, which cannot effectively cope with the challenges of arc discharge faults. It is urgent to propose a new lightweight short-gap arc discharge pattern recognition method to solve the contradiction between the fast requirement for arc discharge pattern recognition in online monitoring and the slow speed of existing recognition methods that cannot be applied to arc discharge faults.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the invention and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0006] The present invention provides a lightweight identification method for arc discharge patterns in oil based on DSCNN. A lightweight deep separable convolutional neural network model is used to mine the time-frequency spectrum of the discharge signal to achieve fast and simple discharge pattern recognition. This solves the contradiction between the requirement for fast arc discharge pattern recognition in online monitoring and the slow speed of existing recognition methods that cannot be applied to arc discharge faults.

[0007] A lightweight identification method for arc discharge patterns in oil based on DSCNN includes: Step 1: Carry out short-gap oil arc discharge test experiment of oil-immersed power transformer; Step 2: Use sensors to collect the discharge time domain data of the arc discharge test experiment, perform wavelet transform on it, and generate a time-frequency spectrum; Step 3: Divide the time-frequency spectrum into a training set and a test set, and use the training set to train a lightweight deep separable convolutional neural network model. The lightweight deep separable convolutional neural network model decomposes the standard convolution operation into two parts: depthwise convolution and pointwise convolution. Regularization and activation functions are added after both depthwise convolution and pointwise convolution to form a DSCNN module. Use the test set to test the trained model and record the first training parameters. Step 4: Use the training set to train a conventional convolutional neural network model, use the test set to detect the trained model, and record the second training parameters. The convolutional neural network model includes convolution, pooling, and full connection operations. Step 5: By comparing the first training parameters and the second training parameters, when the lightweight deep separable convolutional neural network model is used to carry out short-gap oil arc discharge pattern recognition, the parameter scale is smaller, the calculation amount is less, and the inference time is shorter, which confirms the lightweight advantage of the short-gap arc discharge pattern recognition method based on DSCNN and meets the requirements of online monitoring.

[0008] In the lightweight identification method of oil-in-arc discharge patterns based on DSCNN, a short-gap oil-in-arc discharge test experiment for an oil-immersed power transformer is carried out, which includes building an oil-in-arc discharge test platform, wherein an AC power supply is regulated by a voltage regulator to apply a controllable voltage to a closed discharge platform, and the discharge platform has electrodes for discharge and a voltage sensor, a high-frequency coil, an ultrasonic sensor, a pressure sensor, and an ultra-high-frequency sensor for measuring discharge parameters.

[0009] In the lightweight identification method of arc discharge patterns in oil based on DSCNN, the short-gap arc fault models tested by the discharge platform are divided into needle-plate discharge, surface discharge, and inter-turn discharge types.

[0010] In the lightweight identification method of arc discharge patterns in oil based on DSCNN, in step 2, the voltage, current, ultrasound, oil pressure, and ultra-high frequency time domain signals of the arc discharge test experiment are collected, and the corresponding time-frequency spectrum is generated through wavelet transform.

[0011] In the lightweight recognition method for arc discharge patterns in oil based on DSCNN, a training set is used to train a lightweight deep separable convolutional neural network model. The input layer compresses the time-frequency spectrum size to 64 pixels × 64 pixels and then transmits it to the convolution layer. The convolution layer first captures the global structure and preliminary features of the time-frequency spectrum through the convolution kernel, then extracts deep features through four DSCNN modules, and then downsamples through a pooling operation and uses the torch.cat() method to perform tensor splicing. The obtained features are transmitted to the fully connected layer. The extracted features are classified through two fully connected layers, and the classification results are transmitted to the output layer. The output layer uses the Softmax function to output the predicted probability to achieve the recognition task.

[0012] In the lightweight identification method of arc discharge patterns in oil based on DSCNN, the training set trains a conventional convolutional neural network model, and the network structure is composed of an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer connected to each other. The first training parameter and the second training parameter respectively include recognition accuracy, parameter scale, calculation amount, model size, and inference time.

[0013] In the lightweight recognition method for arc discharge patterns in oil based on DSCNN, after training, if the recognition accuracy of the test set is greater than 80%, the training parameters are recorded; if the recognition accuracy of the test set is less than 80%, the learning rate is manually modified and training is repeated.

[0014] In the lightweight identification method of arc discharge patterns in oil based on DSCNN, the algorithm for generating a time-frequency spectrum from a time domain signal through wavelet transform is as follows: ,in, Represents a time domain signal In time and scale The wavelet transform value at is the mother wavelet function, is the complex conjugate of the mother wavelet function, is the scale parameter, is the time shift parameter.

[0015] In the lightweight identification method of arc discharge patterns in oil based on DSCNN, in step 3, the regularization processing selects a batch normalization module to perform normalization and regularization processing.

[0016] In the lightweight identification method of arc discharge patterns in oil based on DSCNN, in step 3, the activation function is the ReLU6 function, and the expression of the ReLU6 function is: , where x is the output value of the activation function, and the max() and min() functions are used to obtain the maximum and minimum values ​​of the sequence respectively.

[0017] In the lightweight identification method of arc discharge patterns in oil based on DSCNN, the ReLU6 function limits the output value and truncates the value greater than 6.

[0018] Compared with the existing technology, the present invention has the following advantages: the present invention achieves lightweight data processing by adopting a deep separable convolutional neural network, introducing regularization, improving activation functions, etc., and while ensuring a high recognition accuracy, it improves the inference speed and reduces the requirements for equipment computing power. It is expected to be promoted in the field of arc discharge fault identification and online monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Various other advantages and benefits of the present invention will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiments below. The accompanying drawings are intended only to illustrate preferred embodiments and are not to be construed as limiting the present invention. It should be understood that the drawings described below are merely examples of the present invention, and that those skilled in the art will be able to derive other drawings from these drawings without inventive effort. Throughout the drawings, identical reference numerals are used to denote identical components.

[0020] In the attached figure: Figure 1 The figure is a flow chart of a lightweight identification method for arc discharge patterns in short-gap oil based on DSCNN. Figure 2 The oil-paper insulation arc discharge test experimental platform designed for the present invention; Figure 3 This is a schematic diagram of the convolution process of a depth-wise separable convolutional neural network of the present invention; Figure 4 Schematic diagram of a DSCNN module of the present invention; Figure 5 This is a flow chart of the lightweight depth-separable convolutional neural network model of the present invention.

[0021] The present invention will be further explained below with reference to the accompanying drawings and embodiments. DETAILED DESCRIPTION

[0022] Specific embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although specific embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0023] It should be noted that certain words are used in the specification and claims to refer to specific components. Those skilled in the art should understand that technicians may use different nouns to refer to the same component. This specification and claims do not use the difference in nouns as a way to distinguish components, but use the difference in the functions of the components as the criterion for distinction. As mentioned throughout the specification and claims, "including" or "comprising" is an open term, so it should be interpreted as "including but not limited to". The subsequent description of the specification is a preferred embodiment of the present invention, but the description is based on the general principles of the specification and is not intended to limit the scope of the invention. The scope of protection of the present invention shall be as defined in the attached claims.

[0024] To facilitate understanding of the embodiments of the present invention, further explanation will be given below using specific embodiments as examples in conjunction with the accompanying drawings, and the accompanying drawings do not constitute a limitation on the embodiments of the present invention.

[0025] like Figures 1 to 5 As shown in FIG, the lightweight identification method of arc discharge patterns in oil based on DSCNN includes the following steps: Step 1: Carry out short-gap oil arc discharge test experiment of oil-immersed power transformer; Step 2: Use sensors to collect the discharge time domain data of the arc discharge test experiment, perform wavelet transform on it, and generate a time-frequency spectrum; Step 3: Divide the time-frequency spectrum into a training set and a test set, and use the training set to train a lightweight deep separable convolutional neural network model. The lightweight deep separable convolutional neural network model decomposes the standard convolution operation into two parts: depthwise convolution and pointwise convolution. Regularization and activation functions are added after both depthwise convolution and pointwise convolution to form a DSCNN module. Use the test set to test the trained model and record the first training parameters. Step 4: Use the training set to train a conventional convolutional neural network model, use the test set to detect the trained model, and record the second training parameters. The convolutional neural network model includes convolution, pooling, and full connection operations. Step 5: By comparing the first training parameters and the second training parameters, when the lightweight deep separable convolutional neural network model is used to carry out short-gap oil arc discharge pattern recognition, the parameter scale is smaller, the calculation amount is less, and the inference time is shorter, which confirms the lightweight advantage of the short-gap arc discharge pattern recognition method based on DSCNN and meets the requirements of online monitoring.

[0026] In a preferred embodiment of the lightweight identification method for oil-in-arc discharge patterns based on DSCNN, conducting a short-gap oil-in-arc discharge test experiment for an oil-immersed power transformer includes building an oil-in-arc discharge test platform, wherein an AC power supply is regulated by a voltage regulator to apply a controllable voltage to a closed discharge platform, and the discharge platform has electrodes for discharge and a voltage sensor, a high-frequency coil, an ultrasonic sensor, a pressure sensor, and an ultra-high-frequency sensor for measuring discharge parameters.

[0027] In a preferred embodiment of the lightweight identification method for arc discharge patterns in oil based on DSCNN, the short-gap arc fault models tested by the discharge platform are divided into needle-plate discharge, surface discharge, and inter-turn discharge types.

[0028] In a preferred embodiment of the lightweight identification method of arc discharge patterns in oil based on DSCNN, in step 2, the voltage, current, ultrasound, oil pressure, and ultra-high frequency time domain signals of the arc discharge test experiment are collected, and the corresponding time-frequency spectrum is generated by wavelet transform.

[0029] In a preferred embodiment of the lightweight identification method of arc discharge patterns in oil based on DSCNN, a training set is used to train a lightweight deep separable convolutional neural network model, and the input layer compresses the time-frequency spectrum size to 64 pixels × 64 pixels and then transmits it to the convolution layer; the convolution layer first captures the global structure and preliminary features of the time-frequency spectrum through the convolution kernel, and then extracts deep features through four DSCNN modules, and then downsamples through the pooling operation, and uses the torch.cat() method to perform tensor splicing; the obtained features are transmitted to the fully connected layer; the extracted features are classified through two fully connected layers, and the classification results are transmitted to the output layer; the output layer uses the Softmax function to output the predicted probability to achieve the recognition task.

[0030] In a preferred embodiment of the lightweight identification method of arc discharge patterns in oil based on DSCNN, the training set trains a conventional convolutional neural network model, and the network structure is composed of an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer connected to each other.

[0031] In a preferred embodiment of the lightweight identification method for arc discharge patterns in oil based on DSCNN, the first training parameter and the second training parameter respectively include recognition accuracy, parameter scale, computational complexity, model size, and inference time.

[0032] In a preferred embodiment of the lightweight identification method for arc discharge patterns in oil based on DSCNN, after training, if the test set recognition accuracy is greater than 80%, the training parameters are recorded; if the test set recognition accuracy is less than 80%, the learning rate is manually modified and retraining is performed.

[0033] In a preferred embodiment of the lightweight identification method of arc discharge patterns in oil based on DSCNN, the algorithm for generating a time-frequency spectrum through wavelet transform of the time domain signal is as follows: ,in, Represents a time domain signal In time and scale The wavelet transform value at is the mother wavelet function, is the complex conjugate of the mother wavelet function, is the scale parameter, is the time shift parameter.

[0034] In a preferred embodiment of the lightweight identification method of arc discharge patterns in oil based on DSCNN, in step 3, the regularization processing selects a batch normalization module to perform normalization and regularization processing.

[0035] In a preferred embodiment of the lightweight identification method of arc discharge patterns in oil based on DSCNN, in step 3, the activation function is a ReLU6 function, and the expression of the ReLU6 function is: .

[0036] Among them, x is the output value of the activation function, and the max() and min() functions are used to obtain the maximum and minimum values ​​of the sequence respectively.

[0037] In a preferred embodiment of the lightweight identification method for arc discharge patterns in oil based on DSCNN, the ReLU6 function limits the output value and truncates the value greater than 6.

[0038] In one embodiment, Figure 1 As shown in Figure 2, the lightweight identification method of arc discharge patterns in oil based on DSCNN includes: Step 1: If Figure 2 As shown, a short gap arc discharge test experiment in oil was carried out; Step 2: Use the sensor to collect the discharge time domain data, perform wavelet transform on the time domain signal, and generate a time-frequency spectrum; Step 3: Divide the time-frequency spectrum into a training set and a test set, use the training set to train a lightweight deep separable convolutional neural network model, use the test set to test the trained model, and record the training parameters; Step 4: Use the same training set as in step 3 to train a conventional convolutional neural network model, use the same test set as in step 3 to test the trained model, and record the training parameters. Step 5: By comparing the training parameters obtained in step 3 and step 4, the lightweight advantage of the short gap arc discharge pattern recognition method based on DSCNN is confirmed.

[0039] In one embodiment, the short gap is in the order of mm.

[0040] In one embodiment, the stochastic gradient descent method is used to solve the weight parameters that minimize the loss value, and the optimal values ​​of the learning rate and iteration round network hyperparameters are automatically selected through cross-validation and grid search methods to ensure that the model converges without taking up too many computing resources.

[0041] In one embodiment, the expression of the ReLU6 function is: The ReLU6 function limits the output value and truncates the value greater than 6, which helps prevent the activation value from being too large and facilitates the deployment of the model on a low-precision platform. Figure 5 As shown in the figure, the lightweight deep separable convolutional neural network model has the following model learning process: compress the time-frequency spectrum size to 64 pixels × 64 pixels; quickly capture the global structure and preliminary features of the time-frequency spectrum through a conventional convolution kernel; further extract deep features through four DSCNN modules; downsample through the pooling layer; use the torch.cat() method to perform tensor splicing; classify the extracted features through two fully connected layers; use the Softmax function to output the predicted probability to achieve the recognition task.

[0042] Specifically, the network structure parameters are shown in the table.

[0043]

[0044] The first recorded training parameters described in step 3 include recognition accuracy, parameter scale, computational effort, model size, and inference time. In step 5, by comparing the training parameters of the lightweight deep separable convolutional neural network model and the conventional convolutional neural network model, it is confirmed that the lightweight deep separable convolutional neural network model has a smaller parameter scale, less computational effort, and shorter inference time when performing short-gap oil arc discharge pattern recognition, thus meeting the requirements of online monitoring.

[0045] Specifically, assuming that the number of channels of the input feature map is M and the size of the convolution kernel is D k , the number of channels of the output feature map is N, then the parameter calculation formula of the depth convolution is:

[0046] The parameter calculation formula of the point-by-point convolution is:

[0047] The total number of parameters of the lightweight depth-separable convolutional neural network model is:

[0048] Under the premise of maintaining the same receptive field and convolution kernel size, the number of parameters required for the conventional convolutional neural network model is:

[0049] The ratio of the total number of parameters of the lightweight depth-wise separable convolutional neural network model to the number of parameters required by the conventional convolutional neural network model can be expressed as:

[0050] Because N, D k are all positive integers, and the ratio Q is a number that is always less than 1.

[0051] Specifically, the values ​​of the remaining training parameters are shown in the table.

[0052]

[0053] While maintaining performance, this invention significantly reduces the number of parameters and model complexity, making it more suitable for resource-constrained environments such as mobile devices or embedded systems. By employing a deep separable convolutional neural network and improving activation functions, the data processing process is lightweight. This improves inference speed while maintaining high recognition accuracy, reducing the computing power requirements of the device. This method is expected to be widely used in the fields of arc discharge fault identification and online monitoring.

[0054] Although the embodiments of the present invention have been described above with reference to the accompanying drawings, the present invention is not limited to the above-mentioned specific embodiments and application fields. The above-mentioned specific embodiments are merely illustrative and instructive, and are not restrictive. A person skilled in the art, guided by this specification and without departing from the scope of protection of the claims of the present invention, may also devise various forms, all of which fall within the scope of protection of the present invention.

Claims

1. A lightweight identification method for arc discharge patterns in oil based on DSCNN, characterized in that: The steps include: Step 1: Carry out the short gap oil arc discharge test experiment of oil-immersed power transformer; Step 2: Use sensors to collect the discharge time domain data of the arc discharge test experiment, perform wavelet transform on it, and generate a time-frequency spectrum; Step 3: Divide the time-frequency spectrum into a training set and a test set, and use the training set to train a lightweight deep separable convolutional neural network model. The lightweight deep separable convolutional neural network model decomposes the standard convolution operation into two parts: deep convolution and point-by-point convolution. Regularization processing and activation functions are added after deep convolution and point-by-point convolution to form a DSCNN module. Use the test set to detect the trained model and record the first training parameters. Step 4: Use the training set to train a conventional convolutional neural network model, use the test set to detect the trained model, and record the second training parameter. The convolutional neural network model includes convolution, pooling, and full connection operations; Step 5: By comparing the first training parameters and the second training parameters, when the lightweight deep separable convolutional neural network model is used to carry out short-gap oil arc discharge pattern recognition, the parameter scale is smaller, the calculation amount is less, and the reasoning time is shorter, which confirms the lightweight advantage of the short-gap arc discharge pattern recognition method based on DSCNN and meets the requirements of online monitoring.

2. The lightweight identification method of arc discharge pattern in oil based on DSCNN according to claim 1 is characterized in that: Preferably, conducting a short-gap oil-in-oil arc discharge test experiment for an oil-immersed power transformer includes building an oil-in-oil arc discharge test platform, wherein an AC power supply is regulated by a voltage regulator to apply a controllable voltage to a closed discharge platform, the discharge platform has electrodes for discharge and a voltage sensor, a high-frequency coil, an ultrasonic sensor, a pressure sensor, and an ultra-high frequency sensor for measuring discharge parameters of the discharge, and the short-gap arc fault models tested by the discharge platform are divided into needle-plate discharge, surface discharge, and turn-to-turn discharge.

3. The lightweight identification method of arc discharge pattern in oil based on DSCNN according to claim 1 is characterized in that: In step 2, the voltage, current, ultrasound, oil pressure, and UHF time domain signals of the arc discharge test experiment are collected, and the corresponding time-frequency spectrum is generated through wavelet transform.

4. The lightweight identification method of arc discharge pattern in oil based on DSCNN according to claim 1 is characterized in that: In the lightweight deep separable convolutional neural network model trained using the training set, the input layer compresses the time-frequency spectrum size to 64 pixels × 64 pixels and then transmits it to the convolution layer; the convolution layer first captures the global structure and preliminary features of the time-frequency spectrum through the convolution kernel, and then extracts deep features through four DSCNN modules, and then downsamples through the pooling operation, and uses the torch.cat() method to perform tensor splicing; the obtained features are transmitted to the fully connected layer; the extracted features are classified through 2 fully connected layers, and the classification results are transmitted to the output layer; the output layer uses the Softmax function to output the predicted probability to achieve the recognition task.

5. The lightweight identification method of arc discharge pattern in oil based on DSCNN according to claim 1 is characterized in that: In the conventional convolutional neural network model trained with the training set, the network structure is composed of an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer connected to each other. The first training parameter and the second training parameter respectively include recognition accuracy, parameter scale, calculation amount, model size and inference time.

6. The lightweight identification method of arc discharge pattern in oil based on DSCNN according to claim 1 is characterized in that: After the training is completed, if the recognition accuracy of the test set is greater than 80%, the training parameters are recorded; if the recognition accuracy of the test set is less than 80%, the learning rate is manually modified and retraining is performed.

7. The lightweight identification method of arc discharge pattern in oil based on DSCNN according to claim 1 is characterized in that: The algorithm for generating time-frequency spectrograms through wavelet transform of time domain signals is as follows: ,in, Represents a time domain signal In time and scale The wavelet transform value at is the mother wavelet function, is the complex conjugate of the mother wavelet function, is the scale parameter, is the time shift parameter.

8. The lightweight identification method of arc discharge pattern in oil based on DSCNN according to claim 1 is characterized in that: In step 3, the regularization process selects the batch normalization module to perform normalization and regularization.

9. The lightweight identification method of arc discharge pattern in oil based on DSCNN according to claim 1 is characterized in that: In step 3, the activation function is the ReLU6 function, and the expression of the ReLU6 function is: , where x is the output value of the activation function, and the max() and min() functions are used to obtain the maximum and minimum values ​​of the sequence respectively.

10. A lightweight identification method for arc discharge patterns in oil based on DSCNN according to claim 9, characterized in that: The ReLU6 function limits the output value and truncates the value greater than 6.

Citation Information

Patent Citations

  • Driver behavior recognition method based on lightweight convolutional neural network

    CN110532878A

  • Main transformer partial discharge identification method

    CN112070104A

  • Power equipment partial discharge type identification method, equipment and storage medium

    CN114970601A

  • Lightweight convolutional network facial emotion recognition method with auxiliary classifier

    CN116580436A

Cited By

  • Electric arc detection method and system based on lightweight neural network model

    CN119579598A