Integrator for detecting partial discharge signal and partial discharge signal detection method

Abstract

The invention relates to the technical field of transformer detection, and discloses an integrator for detecting a partial discharge signal and an integral signal detection method, and the integrator comprises a signal preprocessing module which is used for obtaining and collecting an initial partial discharge signal of a transformer, and preprocessing the initial partial discharge signal to obtain a preprocessed partial discharge signal; the deep learning model training module is used for constructing a deep learning neural network model and training the deep learning neural network model by using the preprocessed partial discharge signal to obtain a trained deep learning neural network model; the integral calculation module is used for taking the preprocessed partial discharge signal as input data of a trained deep learning neural network model, outputting an effective partial discharge signal through the trained deep learning neural network model, and performing integral calculation on the effective partial discharge signal to obtain a real partial discharge signal; and the result output module is used for outputting the real partial discharge signal in the form of a graph.

Description

Technical Field

[0001] The present invention relates to the technical field of transformer detection, and particularly relates to an integrator for detecting partial discharge signals and a partial discharge signal detection method. Background Art

[0002] The partial discharge signal is weak and easily submerged by interference signals. In order to ensure the accuracy of partial discharge signal detection and the reliability of diagnostic results, it is usually necessary to obtain real partial discharge information through anti-interference during detection to judge the insulation condition of the equipment. Partial discharge can cause insulation damage of the transformer and lead to power outages and other accidents. Anti-interference can provide early warning of insulation faults, which is of great significance for ensuring the safe and stable operation of the power system and improving power supply reliability; it can optimize equipment maintenance management based on accurate monitoring results, reduce costs and extend the service life; and the operating environment of the transformer is complex, with many electromagnetic and non-electromagnetic interferences. Especially, the on-line monitoring system needs a strong anti-interference ability to provide effective data support for long-term stable operation in a complex electromagnetic environment. It can be seen that the existing partial discharge signal detection equipment has problems of poor anti-interference ability and low detection accuracy. Summary of the Invention

[0003] The present invention provides an integrator for detecting partial discharge signals and a partial discharge signal detection method to solve the problems of poor anti-interference ability and low detection accuracy of the existing partial discharge signal detection equipment.

[0004] To achieve the above object, the present invention is realized through the following technical solutions: In the first aspect, the present invention provides an integrator for detecting partial discharge signals, and the integrator includes: A signal preprocessing module, configured to acquire the initial partial discharge signal of the collected transformer, preprocess the initial partial discharge signal to obtain a preprocessed partial discharge signal, and send the preprocessed partial discharge signal to the deep learning model training module; A deep learning model training module, configured to construct a deep learning neural network model, and use the preprocessed partial discharge signal to train the deep learning neural network model to obtain a trained deep learning neural network model; An integration calculation module, configured to call the trained deep learning neural network model in the deep learning model training module, use the preprocessed partial discharge signal as the input data of the trained deep learning neural network model, output an effective partial discharge signal through the trained deep learning neural network model, and further configured to perform integration calculation on the effective partial discharge signal to obtain a real partial discharge signal, and send the real partial discharge signal to the result output module; A result output module, configured to output the real partial discharge signal in the form of a graph.

[0005] Optionally, the signal preprocessing module includes: a signal collector, an analog-to-digital converter, and a first signal filter; The signal collector is configured to collect the initial partial discharge signal in the transformer and send the initial partial discharge signal to the analog-to-digital converter; The analog-to-digital converter is configured to perform signal conversion on the initial partial discharge signal, convert it from an analog signal to a digital signal, and send the converted digital signal to the signal filter; The first signal filter is configured to denoise and frequency-filter the digital signal to obtain a preprocessed partial discharge signal.

[0006] Optionally, the training of the deep learning model training module for the deep learning neural network model includes: Constructing a training dataset, a validation set, and a test set by using the preprocessed partial discharge model and the historical preprocessed partial discharge model; Determining the number of layers of the neural network and the number of neurons in each layer of the neural network, and selecting a corresponding machine learning algorithm; After initializing the weights and biases of the neural network, inputting the training set data into the neural network for forward calculation to obtain a predicted output; Calculating the difference between the predicted output and the true label to obtain a loss value, calculating the gradient according to the loss value, and backpropagating the gradient to each layer of the neural network; Updating the weights and biases of the neural network according to the gradient, and evaluating the neural network by using the validation set or the test set. If the evaluation result reaches a predetermined performance index or the number of training times reaches a predetermined number of iterations, the training of the deep learning neural network model can be completed.

[0007] Optionally, the integral calculation module includes a digital signal processor, a second signal filter, and a deep learning accelerator; The digital signal processor is configured to perform integral calculation on the effective partial discharge signal to obtain an integral calculation result, and send the integral calculation result to the signal filter; The second signal filter is configured to denoise the integral calculation result to obtain a true partial discharge signal; The deep learning accelerator is configured to read the effective partial discharge signal and the true partial discharge signal, and send the effective partial discharge signal and the true partial discharge signal to the deep learning model training module for updating the deep learning neural network model.

[0008] Optionally, the integral calculation module further includes: a digital-to-analog converter; The digital-to-analog converter is configured to read the true partial discharge signal and convert the true partial discharge signal from a digital signal to an analog signal for output.

[0009] Optionally, the result output module includes: a graphic display; The image display is configured to generate corresponding image data according to the true partial discharge signal and display the image data.

[0010] In a second aspect, an embodiment of the present application provides a partial discharge signal detection method, which is applied to an integrator for detecting partial discharge signals in the first aspect. The method includes: Obtain an initial partial discharge signal of a collecting transformer, and preprocess the initial partial discharge signal to obtain a preprocessed partial discharge signal; Construct a deep learning neural network model, and use the preprocessed partial discharge signal to train the deep learning neural network model to obtain a trained deep learning neural network model; Take the preprocessed partial discharge signal as input data of the trained deep learning neural network model, output an effective partial discharge signal through the trained deep learning neural network model, and perform integral calculation on the effective partial discharge signal to obtain a true partial discharge signal; Output the true partial discharge signal in the form of a graph.

[0011] Optionally, the constructing a deep learning neural network model and using the preprocessed partial discharge signal to train the deep learning neural network model to obtain a trained deep learning neural network model includes: Use the preprocessed partial discharge model and the historical preprocessed partial discharge model to construct a training dataset, a validation set, and a test set; Determine the number of layers of the neural network and the number of neurons in each layer of the neural network, and select a corresponding machine learning algorithm; After initializing the weights and biases of the neural network, input the training set data into the neural network for forward calculation to obtain a predicted output; Calculate the difference between the predicted output and the true label to obtain a loss value, calculate the gradient according to the loss value, and backpropagate the gradient to each layer of the neural network; Update the weights and biases of the neural network according to the gradient, and evaluate the neural network using the validation set or the test set. If the evaluation result reaches a predetermined performance index or the number of training times reaches a predetermined number of iterations, the training of the deep learning neural network model can be completed.

[0012] Beneficial effects: The integrator for detecting partial discharge signals provided by the present invention processes signals by receiving signals from partial discharge detection devices and utilizing its built-in integration function, thereby effectively extracting the intensity and frequency characteristics of transformer partial discharge events. By performing integration operations on partial discharge signals, high-frequency noise can be suppressed, interference signals can be filtered out, the signal-to-noise ratio of the signals can be improved, and thus the detection sensitivity and accuracy of partial discharge can be enhanced. Description of the Drawings

[0013] Figure 1 It is a schematic structural diagram of an integrator for detecting partial discharge signals according to a preferred embodiment of the present invention; Figure 2 It is a training flow chart of a deep learning neural network model according to a preferred embodiment of the present invention; Figure 3 It is a flow chart of a partial discharge signal detection method according to a preferred embodiment of the present invention. Detailed Embodiments

[0014] The technical solutions of the present invention will be described clearly and completely below. Apparently, the described embodiments are only a part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0015] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the ordinary meanings understood by those of ordinary skill in the art to which the present invention belongs. The "first", "second" and similar terms used in the present invention do not denote any order, quantity or importance, but are only used to distinguish different components. Similarly, terms such as "a" or "one" do not denote a quantity limitation, but mean that there is at least one. "Connection" or "connected" and similar terms are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right", etc. are only used to represent relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship also changes accordingly.

[0016] Please refer to Figure 1 , an embodiment of the present application provides an integrator for detecting partial discharge signals, and the integrator includes: A signal preprocessing module, configured to obtain an initial partial discharge signal of a collected transformer, preprocess the initial partial discharge signal to obtain a preprocessed partial discharge signal, and send the preprocessed partial discharge signal to a deep learning model training module; A deep learning model training module for constructing a deep learning neural network model and training the deep learning neural network model using the preprocessed partial discharge signals to obtain a trained deep learning neural network model; An integration calculation module for calling the trained deep learning neural network model in the deep learning model training module, using the preprocessed partial discharge signals as input data of the trained deep learning neural network model, outputting effective partial discharge signals through the trained deep learning neural network model, further performing integration calculation on the effective partial discharge signals to obtain real partial discharge signals, and sending the real partial discharge signals to the result output module; A result output module for outputting the real partial discharge signals in the form of a graph.

[0017] In the above embodiment, the integrator design of the present invention can be divided into the following modules: A signal preprocessing module: responsible for sampling and preprocessing of input signals; This module is responsible for receiving external signals and performing necessary sampling and preprocessing operations. It includes an analog-to-digital converter (ADC) to convert analog signals into digital signals, and a signal filter for removing noise and unwanted frequency components.

[0018] A deep learning model training module (DNN): for training the neural network model of the integrator; This module is the core of the entire system. It contains a deep learning neural network (DNN), which is trained through machine learning algorithms. During the training process, a large number of complex signals are input to learn how to distinguish effective signals from noise signals and perform screening, and only the effective signals are transmitted to the next module.

[0019] An integration calculation module: performing integration calculation based on the trained model; After the deep learning model training is completed, this module uses the trained neural network to perform integration calculation on the effective input signals. This module includes a digital signal processor (DSP) or a dedicated deep learning accelerator for performing the forward propagation of the network.

[0020] A result output module: outputting the calculation results to a control system or other applications; After the integration calculation is completed, the result output module converts the processed digital signals back into analog signals or outputs them in digital form for use in subsequent control systems or other applications.

[0021] Optionally, the signal preprocessing module includes: a signal collector, an analog-to-digital converter, and a first signal filter; The signal collector is used to collect the initial partial discharge signal in the transformer and send the initial partial discharge signal to the analog-to-digital converter; The analog-to-digital converter is used to perform signal conversion on the initial partial discharge signal, convert it from an analog signal to a digital signal, and send the converted digital signal to the signal filter; The first signal filter is used to denoise and filter the frequency of the digital signal to obtain a preprocessed partial discharge signal.

[0022] In the above embodiment, the partial discharge signal of the transformer is collected in real time through a high-frequency current sensor or a detection device. The signal usually exists in the form of current or voltage waveforms and may include various noise sources, such as high-frequency noise, environmental interference, etc. Before the deep learning model training, signal preprocessing is carried out. First, the original electrical signal is denoised, and traditional filtering methods are used to remove high-frequency noise, environmental interference, and other irrelevant signals. The key feature parameters of the partial discharge signal of the transformer (such as the amplitude and frequency of the signal) are extracted through time-domain analysis and frequency-domain analysis, and are standardized into the format for inputting into the neural network.

[0023] Optionally, the training of the deep learning neural network model by the deep learning model training module includes: Constructing a training dataset, a validation set, and a test set by using the preprocessed partial discharge model and the historical preprocessed partial discharge model; Determining the number of layers of the neural network and the number of neurons in each layer of the neural network, and selecting the corresponding machine learning algorithm; After initializing the weights and biases of the neural network, inputting the training set data into the neural network for forward calculation to obtain the predicted output; Calculating the difference between the predicted output and the true label to obtain the loss value, calculating the gradient according to the loss value, and backpropagating the gradient to each layer of the neural network; Updating the weights and biases of the neural network according to the gradient, and evaluating the neural network by using the validation set or the test set. If the evaluation result reaches the predetermined performance index or the number of training times reaches the predetermined number of iterations, the training of the deep learning neural network model can be completed.

[0024] In the above embodiment, the specific implementation flowchart is as Figure 2 .

[0025] Start: Initiate the deep learning model training process; Dataset construction: Use a large number of complex signal inputs to construct a training dataset and learn how to distinguish valid signals and noise signals; Define the neural network structure: Determine the number of layers of the neural network, the number of neurons in each layer, etc.; Select a training algorithm: Select a suitable machine learning algorithm, such as gradient descent, Adam, etc.; Initialize network parameters: Initialize the weights and biases of the neural network; Forward propagation: The input data undergoes forward calculations through the neural network to obtain the predicted output; Loss function calculation: Calculate the difference between the predicted output and the true label to obtain the loss value; Backward propagation: Calculate the gradient based on the loss value and backpropagate it to each layer of the network; Parameter update: Update the weights and biases of the network according to the gradient to optimize the network performance; Model evaluation: Use the validation set or test set to evaluate the recognition ability and accuracy of the model; Whether the termination condition is met: Determine whether the model has reached the predetermined performance metrics or the number of training iterations; Model training completed: If the termination condition is met, the training is completed; Model deployment: Deploy the trained model to the integral calculation module for the integral calculation of effective signals.

[0026] Using the collected partial discharge signal sample data, train a neural network model based on a deep learning algorithm. The model uses a convolutional neural network (CNN) and optimizes the parameters of the neural network through the backpropagation algorithm. During the training process, the neural network continuously learns various complex patterns and features of partial discharge signals, so as to effectively distinguish normal operating signals and abnormal partial discharge signals in the transformer. The trained neural network model can classify and identify newly input partial discharge signals, dividing them into different categories, such as: normal discharge, mild discharge, severe discharge. In this way, the model can automatically identify abnormal fluctuations in the signal and accurately distinguish different types of partial discharge events.

[0027] Optionally, the integral calculation module includes a digital signal processor, a second signal filter, and a deep learning accelerator; The digital signal processor is used to perform integral calculation on the effective partial discharge signal to obtain the integral calculation result and send the integral calculation result to the signal filter; The second signal filter is used to denoise the integral calculation result to obtain the true partial discharge signal; The deep learning accelerator is used to read the effective partial discharge signal and the true partial discharge signal and send the effective partial discharge signal and the true partial discharge signal to the deep learning model training module for updating the deep learning neural network model.

[0028] In the above embodiments, after the partial discharge signal is identified, the system performs an integration operation on it through an integration algorithm. After the output signal of the deep learning model is integrated, the accuracy and processing precision of the signal can be improved. The integrated signal will be further enhanced to remove noise and unnecessary interference, making the final signal more reliable and facilitating subsequent analysis and monitoring.

[0029] Optionally, the integration calculation module further includes: a digital-to-analog converter; The digital-to-analog converter is used to read the real partial discharge signal and convert the real partial discharge signal from a digital signal to an analog signal for output.

[0030] Optionally, the result output module includes: a graphic display; The image display is used to generate corresponding image data according to the real partial discharge signal and display the image data.

[0031] In the above embodiments, all signal processing results will be output in real time in the system, and the operation and maintenance personnel can view the change trend and its status of the partial discharge signal through the graphic interface.

[0032] As Figure 3 shown, an embodiment of the present application further provides a method for detecting partial discharge signals, which is applied to an integrator for detecting partial discharge signals as described above. The method includes: Obtain the initial partial discharge signal of the acquisition transformer, and preprocess the initial partial discharge signal to obtain a preprocessed partial discharge signal; Construct a deep learning neural network model, and use the preprocessed partial discharge signal to train the deep learning neural network model to obtain a trained deep learning neural network model; Use the preprocessed partial discharge signal as the input data of the trained deep learning neural network model, output an effective partial discharge signal through the trained deep learning neural network model, and perform an integration calculation on the effective partial discharge signal to obtain a real partial discharge signal; Output the real partial discharge signal in the form of a graph.

[0033] Optionally, the constructing a deep learning neural network model and using the preprocessed partial discharge signal to train the deep learning neural network model to obtain a trained deep learning neural network model includes: Use the preprocessed partial discharge model and the historical preprocessed partial discharge model to construct a training dataset, a validation dataset, and a test dataset; Determine the number of layers of the neural network and the number of neurons in each layer of the neural network, and select the corresponding machine learning algorithm; After initializing the weights and biases of the neural network, the training set data is input into the neural network for forward calculation to obtain the predicted output; Calculate the difference between the predicted output and the true label to obtain the loss value, calculate the gradient based on the loss value, and backpropagate the gradient to each layer of the neural network; Update the weights and biases of the neural network according to the gradient, and evaluate the neural network using the validation set or test set. If the evaluation result reaches the predetermined performance index or the number of training times reaches the predetermined number of iterations, the training of the deep learning neural network model can be completed.

[0034] The preferred specific embodiments of the present invention have been described in detail above. It should be understood that those of ordinary skill in the art can make many modifications and variations based on the concept of the present invention without creative labor. Therefore, all technical solutions that can be obtained by those skilled in the art in the technical field of the present invention based on the concept of the present invention through logical analysis, reasoning, or limited experiments on the basis of the prior art should be within the protection scope determined by the claims.

Claims

1. An integrator for detecting partial discharge signals, characterized in that: The integrator comprises: A signal preprocessing module is used to acquire an initial partial discharge signal of the acquisition transformer, preprocess the initial partial discharge signal to obtain a preprocessed partial discharge signal, and send the preprocessed partial discharge signal to a deep learning model training module; A deep learning model training module is used to construct a deep learning neural network model, and use the pre-processed partial discharge signal to train the deep learning neural network model to obtain a trained deep learning neural network model; An integral calculation module is used to call the deep learning neural network model trained in the deep learning model training module, use the pre-processed partial discharge signal as input data of the trained deep learning neural network model, output an effective partial discharge signal through the trained deep learning neural network model, and perform integral calculation on the effective partial discharge signal to obtain a real partial discharge signal, and send the real partial discharge signal to the result output module; The result output module is used to output the real partial discharge signal in the form of a graph.

2. The integrator for detecting partial discharge signals according to claim 1, characterized in that: The signal preprocessing module includes: a signal collector, an analog-to-digital converter and a first signal filter; The signal collector is used to collect the initial partial discharge signal in the transformer and send the initial partial discharge signal to the analog-to-digital converter; The analog-to-digital converter is used to perform signal conversion on the initial partial discharge signal, converting it from an analog signal to a digital signal, and sending the converted digital signal to a signal filter; The first signal filter is used to perform denoising and frequency filtering on the digital signal to obtain a pre-processed partial discharge signal.

3. The integrator for detecting partial discharge signals according to claim 1, characterized in that: The deep learning model training module trains the deep learning neural network model including: Using the preprocessed partial discharge model and the historical preprocessed partial discharge model to construct a training data set, a validation set and a test set; Determine the number of neural network layers and the number of neurons in each layer, and select the corresponding machine learning algorithm; After initializing the weights and biases of the neural network, the training set data is input into the neural network for forward calculation to obtain the predicted output; Calculate the difference between the predicted output and the true label to get the loss value, calculate the gradient based on the loss value, and backpropagate the gradient to each layer of the neural network; The weights and biases of the neural network are updated according to the gradient, and the neural network is evaluated using a validation set or a test set. If the evaluation result reaches the predetermined performance index or the number of training iterations reaches the predetermined number, the training of the deep learning neural network model can be completed.

4. The integrator for detecting partial discharge signals according to claim 1, characterized in that: The integral calculation module includes a digital signal processor, a second signal filter and a deep learning accelerator; The digital signal processor is used to perform integration calculation on the effective partial discharge signal to obtain an integration calculation result, and send the integration calculation result to the signal filter; The second signal filter is used to denoise the integral calculation result to obtain a true partial discharge signal; The deep learning accelerator is used to read the effective partial discharge signal and the real partial discharge signal, and send the effective partial discharge signal and the real partial discharge signal to the deep learning model training module for updating the deep learning neural network model.

5. The integrator for detecting partial discharge signals according to claim 1, characterized in that: The integral calculation module also includes: a digital-to-analog converter; The digital-to-analog converter is used to read the real partial discharge signal, and convert the real partial discharge signal from a digital signal into an analog signal for output.

6. The integrator for detecting partial discharge signals according to claim 1, characterized in that: The result output module includes: a graphic display; The image display is used to generate corresponding image data according to the real partial discharge signal and display the image data.

7. A partial discharge signal detection method, applied to an integrator for detecting partial discharge signals as claimed in claims 1 to 6, characterized in that: The method comprises: Acquire an initial partial discharge signal of the acquisition transformer, and preprocess the initial partial discharge signal to obtain a preprocessed partial discharge signal; Constructing a deep learning neural network model, and using the pre-processed partial discharge signal to train the deep learning neural network model to obtain a trained deep learning neural network model; The pre-processed partial discharge signal is used as input data of a trained deep learning neural network model, an effective partial discharge signal is output through the trained deep learning neural network model, and an integral calculation is performed on the effective partial discharge signal to obtain a real partial discharge signal; The actual partial discharge signal is output in the form of a graph.

8. The method for detecting partial discharge signals according to claim 7, characterized in that: The deep learning neural network model is constructed, and the deep learning neural network model is trained by using the pre-processed partial discharge signal to obtain a trained deep learning neural network model, including: Using the preprocessed partial discharge model and the historical preprocessed partial discharge model to construct a training data set, a validation set and a test set; Determine the number of neural network layers and the number of neurons in each layer, and select the corresponding machine learning algorithm; After initializing the weights and biases of the neural network, the training set data is input into the neural network for forward calculation to obtain the predicted output; Calculate the difference between the predicted output and the true label to get the loss value, calculate the gradient based on the loss value, and backpropagate the gradient to each layer of the neural network; The weights and biases of the neural network are updated according to the gradient, and the neural network is evaluated using a validation set or a test set. If the evaluation result reaches the predetermined performance index or the number of training iterations reaches the predetermined number, the training of the deep learning neural network model can be completed.

Images