Transformer discharge detection method and system with anti-interference function

Abstract

The invention belongs to the technical field of power equipment state monitoring, and discloses a transformer discharge detection method and system with an anti-interference function, and the method comprises the steps: obtaining a multi-mode discharge original signal of a transformer, carrying out the synchronization of a coefficient through a channel, anchoring a synchronous interference signal, extracting initial multi-dimensional features, and carrying out the detection of the multi-mode discharge original signal. Performing weighted average fusion on the initial multi-dimensional features according to a channel synchronization coefficient to obtain an anti-interference feature vector; constructing a feature screening-interference immune lightweight neural network model, inputting anti-interference features, taking a channel synchronization coefficient as an interference immune factor, and outputting a discharge type identification result; and obtaining an evaluation index by combining the peak amplitude of the multi-mode discharge original signal of the transformer after the de-noising preprocessing, and obtaining a discharge risk level based on the evaluation index. According to the invention, accurate identification and risk grade evaluation of transformer discharge are realized, and the problems of poor anti-interference performance and the like are solved.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition monitoring technology, and in particular to a transformer discharge detection method and system with anti-interference function. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Transformers are core equipment in power systems, and their operating status directly affects the safe and stable operation of the power grid. Partial discharge is an important precursor to transformer insulation degradation. Timely and accurate detection of transformer partial discharge types and assessment of discharge severity can effectively prevent transformer failures and reduce power system operation and maintenance costs. Therefore, transformer partial discharge detection technology has always been a research focus in the field of power equipment condition monitoring.

[0004] In existing technologies, detection schemes for multimodal discharge signals of transformers typically utilize UHF, AE, and TEV multi-channel sensors to acquire raw multimodal signals during transformer operation. Then, traditional algorithms such as filtering and wavelet denoising are used to preprocess the raw signals to remove environmental interference and equipment noise. Next, feature parameters such as peak value, kurtosis, and pulse width are extracted from the denoised signals. Finally, the extracted features are input into machine learning models such as convolutional neural networks to achieve discharge type identification and severity assessment.

[0005] However, the study found that existing technologies have many shortcomings in practical applications, including: Existing noise reduction algorithms lack specificity and can only remove noise in fixed frequency bands. They cannot effectively identify and filter out synchronous interference signals that are similar to the characteristics of discharge signals, resulting in a large number of interference components remaining in the signal after noise reduction, which affects the accuracy of subsequent feature extraction. In the feature fusion stage, existing technologies typically involve simply splicing together features from multiple channels, which can lead to the features corresponding to interference signals misleading the identification results and further reducing the reliability of discharge type identification and severity assessment. In existing technologies, feature extraction and model recognition are independent of each other. During the noise reduction process, parameters related to interference features are not retained, which makes it impossible for the model to adaptively adjust the recognition strategy according to the interference intensity. When faced with complex interference scenarios, the recognition accuracy drops significantly. Furthermore, the existing recognition network model has a complex structure, a large number of parameters, and a long training cycle. It also lacks lightweight design for the characteristics of multimodal discharge signals, making it difficult to meet the needs of real-time on-site detection of power equipment. Summary of the Invention

[0006] To address the aforementioned issues, this invention proposes a transformer discharge detection method with anti-interference capabilities. By employing synchronous interference anchoring noise reduction, weighted feature fusion, and a lightweight interference immunity model, it achieves accurate identification of transformer discharge and risk level assessment, thus resolving problems such as poor anti-interference performance.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a transformer discharge detection method with anti-interference function, comprising the following steps: The original signal of multimode discharge of transformer is obtained, and noise reduction preprocessing is performed on the original signal of multimode discharge of transformer by calculating the channel synchronization coefficient and anchoring the synchronization interference signal. Based on the original multi-mode discharge signal of the transformer after noise reduction and preprocessing, initial multi-dimensional features are extracted, and the initial multi-dimensional features are weighted and averaged according to the channel synchronization coefficient to obtain the anti-interference feature vector. A lightweight neural network model is constructed and an interference immune layer is added to obtain a feature selection-interference immune lightweight neural network model. The anti-interference feature vector is input, the channel synchronization coefficient is used as the interference immune factor, and the discharge type identification result is output. Based on the discharge type identification results, an evaluation index is obtained by combining the peak amplitude of the original multi-mode discharge signal of the transformer after noise reduction preprocessing, and the discharge risk level is obtained based on the evaluation index.

[0008] As an optional implementation, the original multi-mode discharge signal of the transformer includes electromagnetic pulse signal, vibration signal and transient voltage signal; the noise reduction preprocessing is specifically as follows: based on the original multi-mode discharge signal of the transformer, the channel synchronization coefficient of the three channels is first calculated, the synchronization interference signal is anchored according to the channel synchronization coefficient, the channel with the weakest interference signal intensity is selected as the reference channel, and the signals of the other two channels are differentially calculated with the reference channel signal to obtain the preprocessed original multi-mode discharge signal of the transformer.

[0009] As an alternative implementation method, the channel synchronization coefficient is calculated as follows: ; in, The amplitude of the electromagnetic pulse signal. The amplitude of the vibration signal. The amplitude of the transient voltage signal. The average value of the electromagnetic pulse signal amplitude at all times. The average value of the vibration signal amplitude at all times. The average value of the transient voltage signal amplitude over all time points. Total sampling time The sampling time.

[0010] As an alternative implementation, the initial multidimensional features are weighted and averaged based on the channel synchronization coefficient, as follows: ;in, For the first Dimensional fusion features For the first The channel synchronization coefficient corresponding to the dimensional feature. For the first Initial multidimensional features; Arrange all fused features in a fixed order to obtain the anti-interference feature vector.

[0011] As an alternative implementation, the feature selection-interference immunity lightweight neural network model sequentially includes an input layer, a convolutional layer, an interference immunity layer, and an output layer. An anti-interference feature vector is input into the input layer, normalized, and then output to the convolutional layer to extract local key features. An activation function is then applied after the convolutional layer, followed by pooling, and the output convolutional feature vector is sent to the interference immunity layer. The channel synchronization coefficient is used as the interference immunity factor to weight and correct the convolutional feature vector, and the interference-immunized feature vector is output to the output layer. The output layer adopts a fully connected layer structure and outputs the discharge type identification result.

[0012] As an alternative implementation, the formula for weighting and correcting the convolutional feature vector is: ;in, To interfere with the characteristics of post-immunization, The convolutional feature vector This is the channel synchronization coefficient.

[0013] As an alternative implementation method, the formula for calculating the evaluation index is: ;in, To evaluate the index, This represents the normalized peak amplitude of the original multimode discharge signal from the transformer after noise reduction preprocessing. This is the weighting coefficient for the discharge type.

[0014] Secondly, the present invention provides a transformer discharge detection system with anti-interference function, comprising the following modules: The data acquisition and preprocessing module is configured to: acquire the original signal of transformer multimode discharge, calculate the channel synchronization coefficient and anchor the synchronization interference signal, and perform noise reduction preprocessing on the original signal of transformer multimode discharge; The feature extraction module is configured to: extract initial multi-dimensional features based on the original multi-mode discharge signal of the transformer after noise reduction preprocessing; and perform weighted average fusion of the initial multi-dimensional features according to the channel synchronization coefficient to obtain an anti-interference feature vector. The identification and detection module is configured to: construct a lightweight neural network model and add an interference immune layer to obtain a feature screening-interference immune lightweight neural network model, input an anti-interference feature vector, use the channel synchronization coefficient as the interference immune factor, and output the discharge type identification result; The evaluation module is configured to: obtain an evaluation index based on the discharge type identification result and the peak amplitude of the original multi-mode discharge signal of the transformer after noise reduction preprocessing; and obtain the discharge risk level based on the evaluation index.

[0015] Thirdly, the present invention provides an electronic device, including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the transformer discharge detection method with anti-interference function described in the first aspect.

[0016] Fourthly, the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the transformer discharge detection method with anti-interference function described in the first aspect.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention anchors the synchronization interference signal by calculating the channel synchronization coefficient, and performs noise reduction preprocessing on the original signal of transformer multimode discharge. It can accurately target and remove synchronization interference components with similar characteristics to the discharge signal, greatly reduce the interference residue of the signal after noise reduction, and significantly improve the accuracy of subsequent feature extraction.

[0018] This invention utilizes channel synchronization coefficients to perform weighted average fusion of initial multidimensional features, thereby reducing the weight of features corresponding to interference signals and increasing the weight of effective discharge features. The constructed anti-interference feature vector can accurately characterize the core characteristics of discharge, avoid misleading identification and assessment by interference features, and improve the reliability of discharge type identification and severity assessment.

[0019] This invention adds an interference immunity layer to a lightweight neural network and introduces the channel synchronization coefficient as an interference immunity factor into the model, so that the model can adaptively correct the convolution features according to the interference intensity. The stronger the interference, the more obvious the suppression effect on redundant features, thus realizing an anti-interference closed loop in the feature processing and recognition process, and greatly improving the recognition accuracy in complex interference scenarios.

[0020] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0021] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0022] Figure 1 The above is a flowchart illustrating the framework of a transformer discharge detection method with anti-interference function provided in Embodiment 1 of the present invention. Detailed Implementation

[0023] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0024] It should be noted that the following detailed description is exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0025] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0026] Example 1 like Figure 1 As shown, this embodiment provides a transformer discharge detection method with anti-interference function for the online partial discharge detection scenario of a 220kV oil-immersed power transformer, including the following steps: S1: Obtain the original signal of transformer multimode discharge, calculate the channel synchronization coefficient and anchor the synchronization interference signal to perform noise reduction preprocessing on the original signal of transformer multimode discharge.

[0027] In step S1, three types of sensors are deployed in the transformer tank and winding area to achieve synchronous data acquisition. Specifically, three types of sensors are deployed in the transformer tank and winding area: Ultra-high frequency (UHF) electromagnetic pulse sensor: center frequency of 400MHz, sampling rate of 2GS / s, installed in the UHF detection window reserved in the transformer box; Acoustic emission (AE) vibration sensor: resonant frequency of 200kHz, sampling rate of 2MHz, magnetically fixed to the lower middle part of the side wall of the enclosure; Transient ground voltage (TEV) sensor: bandwidth 5-120MHz, sampling rate 1GS / s, attached to the metal casing of the transformer grounding terminal. The sampling trigger times of the three types of sensors are completely synchronized, with a total sampling duration T=20s, and the sampling time t takes values ​​of 1, 2, ..., 4× .

[0028] Based on the three types of sensors mentioned above, the acquired raw signals of transformer multimode discharge include electromagnetic pulse signals, vibration signals, and transient voltage signals, and the amplitude of the electromagnetic pulse signal under transformer operating conditions is further obtained. Amplitude of vibration signal and the amplitude of the transient voltage signal This forms a three-channel multimodal discharge raw signal dataset.

[0029] In step S1, the noise reduction preprocessing is based on the original signal of transformer multimode discharge. First, the channel synchronization coefficient of the three-channel signal is calculated. The synchronization interference signal is anchored according to the channel synchronization coefficient. Then, the channel with the weakest interference signal strength is selected as the reference channel. The signals of the other two channels are differentially processed with the reference channel signal to obtain the preprocessed original signal of transformer multimode discharge.

[0030] Among them, the channel synchronization coefficient ( The closer the value is to 1, the stronger the three-channel synchronization interference. The calculation method is as follows: ; in, The amplitude of the electromagnetic pulse signal. The amplitude of the vibration signal. The amplitude of the transient voltage signal. The average value of the electromagnetic pulse signal amplitude at all times. The average value of the vibration signal amplitude at all times. The average value of the transient voltage signal amplitude over all time points. Total sampling time The sampling time.

[0031] based on After anchoring the synchronization interference signal, in this embodiment, the TEV channel with the weakest interference signal strength is selected as the reference channel. The UHF and AE channel signals are respectively differentially analyzed with the TEV channel signal, as shown below: ; Finally, the preprocessed original multimode discharge signal of the transformer is obtained, including the noise-reduced electromagnetic pulse signal. Vibration signals And the transient voltage signal used as a reference.

[0032] With channel synchronization coefficient To target and synchronize interference, this method overcomes the limitations of traditional filtering which only removes noise in fixed frequency bands, enabling accurate identification of interference with similar characteristics to discharge signals. Simultaneously, it employs a denoising method combining the weakest channel reference with differential operations to achieve targeted cancellation of interference signals. After denoising, the signal-to-noise ratio is improved to 28dB, which is 15dB higher than traditional wavelet denoising algorithms, significantly enhancing the effectiveness of subsequent feature extraction.

[0033] S2: Based on the original multi-mode discharge signal of the transformer after noise reduction preprocessing, the initial multi-dimensional features are extracted, and the initial multi-dimensional features are weighted and averaged according to the channel synchronization coefficient to obtain the anti-interference feature vector.

[0034] In step S2, for the denoised three-channel signal (the original multi-mode discharge signal of the transformer after denoising preprocessing), three core features—peak value, kurtosis, and discharge pulse width—are extracted, resulting in a total of 9 initial features. ( =1,2,...,9), the features are defined as follows: Peak value: The maximum amplitude of a single-channel signal within the sampling period, reflecting the intensity of the discharge pulse; Kurtosis: The steepness of the probability distribution of signal amplitude. The higher the kurtosis value, the stronger the sparsity of the discharge pulse. Discharge pulse width: The duration of the pulse signal from the rising edge threshold (10% peak value) to the falling edge threshold (10% peak value), reflecting the duration of the discharge.

[0035] The initial multidimensional features are weighted and averaged based on the channel synchronization coefficient, and expressed as follows: ;in, For the first Dimensional fusion features For the first The channel synchronization coefficient corresponding to the dimensional feature. For the first Initial multidimensional features; in this embodiment, the UHF, AE, and TEV channels... The values ​​are 0.93, 0.91, and 0.89, respectively.

[0036] All fused features are arranged in a fixed order to obtain an anti-interference feature vector; specifically, the fused 9-dimensional features are arranged in a fixed order of "UHF-peak value, UHF-kurtosis, UHF-pulse width, AE-peak value...TEV-pulse width" to obtain a 9-dimensional anti-interference feature vector. .

[0037] By introducing a channel synchronization coefficient as a feature fusion weight factor, the feature weight corresponding to the interference signal is reduced, and the proportion of effective discharge feature weight is increased, thus solving the problem of interference misleading caused by simple splicing of traditional features. At the same time, the anti-interference feature vector constructed in this example improves the representation accuracy of discharge characteristics by 35%, and the correlation between features and discharge type is significantly enhanced compared with the unweighted feature vector.

[0038] S3: Construct a lightweight neural network model and add an interference immune layer to obtain a feature-selection-interference immune lightweight neural network model. Input anti-interference features, use the channel synchronization coefficient as the interference immune factor, and output the discharge type identification result.

[0039] In step S3, the constructed feature selection-interference immune lightweight neural network model sequentially includes an input layer, a convolutional layer, an interference immune layer, and an output layer; specifically: Input layer: Input 9-dimensional anti-interference feature vector The min-max normalization process is performed to eliminate the difference in feature dimensions, and the normalized feature vector is output to the convolutional layer. Convolutional layers: used to extract key local features and reduce dimensionality; each convolutional layer has 20 one-dimensional convolutional kernels with a kernel size of 3 and a stride of 1. A ReLU activation function is applied after the convolutional layer to introduce a non-linear mapping, and finally a max-pooling layer (with a kernel size of 2) is applied to perform pooling operations, outputting a 20-dimensional convolutional feature vector. To interfere with the immune layer.

[0040] Interference Immune Layer: Using the channel synchronization coefficient as an interference immune factor, the convolutional feature vector is weighted and corrected, and the interference-immunized feature vector is output to the output layer. The formula for weighting and correcting the convolutional feature vector is as follows: ;in, To interfere with the characteristics of post-immunization, The convolutional feature vector This is the channel synchronization coefficient. In this embodiment, .

[0041] Output layer: Employs a fully connected layer structure to integrate the 20-dimensional features after interference immunity. Mapped to a 4-dimensional output space (corresponding to 4 typical discharge types: air gap discharge, surface discharge, floating potential discharge, and metal particle discharge), and connected to the Softmax activation function, the probability distribution vector of each discharge type is output.

[0042] For example, the model is trained using a standard sample set of transformer partial discharge (containing 1000 samples of each of 4 discharge types), with a training batch size of 64 and 150 iterations. After the model converges, the anti-interference feature vector of this embodiment is input. The output probability distribution vector is [0.01, 0.03, 0.95, 0.01], and the discharge type is determined to be floating potential discharge.

[0043] S4: Based on the discharge type identification results, and combined with the peak amplitude of the original multi-mode discharge signal of the transformer after noise reduction preprocessing, an evaluation index is obtained, and the discharge risk level is obtained based on the evaluation index.

[0044] In step S4, the peak amplitude of the original multi-mode discharge signal of the transformer after noise reduction preprocessing is calculated (the peak amplitude of the three-channel signal after noise reduction). , , The maximum value among the three is taken as the total peak amplitude. The peak amplitude is normalized to the [0,1] interval. In this embodiment, the normalized peak amplitude is... .

[0045] The formula for calculating the evaluation index is: ;in, To evaluate the index, This represents the normalized peak amplitude of the original multimode discharge signal from the transformer after noise reduction preprocessing. The weighting coefficients are for different discharge types. In this embodiment, the weighting coefficient for air gap discharge is 0.70, the weighting coefficient for surface discharge is 0.80, the weighting coefficient for floating potential discharge is 0.85, and the weighting coefficient for metal particle discharge is 0.90.

[0046] Based on the actual needs of power operation and maintenance, the correspondence between assessment indices and risk levels is established: <0.3: Low risk (weak discharge, equipment can continue to operate, monthly monitoring recommended); 0.3≤ <0.6: Medium risk (slow discharge development, weekly monitoring and trend tracking recommended); ≥0.6: High risk (active discharge, risk of insulation breakdown, requires immediate shutdown and maintenance).

[0047] Example 2 This embodiment provides a transformer discharge detection system with anti-interference function, including the following modules: The data acquisition and preprocessing module is configured to: acquire the original signal of transformer multimode discharge, calculate the channel synchronization coefficient and anchor the synchronization interference signal, and perform noise reduction preprocessing on the original signal of transformer multimode discharge; The feature extraction module is configured to: extract initial multi-dimensional features based on the original multi-mode discharge signal of the transformer after noise reduction preprocessing; and perform weighted average fusion of the initial multi-dimensional features according to the channel synchronization coefficient to obtain an anti-interference feature vector. The identification and detection module is configured to: construct a lightweight neural network model and add an interference immune layer to obtain a feature screening-interference immune lightweight neural network model, input anti-interference features, use the channel synchronization coefficient as the interference immune factor, and output the discharge type identification result; The evaluation module is configured to: obtain an evaluation index based on the discharge type identification result and the peak amplitude of the original multi-mode discharge signal of the transformer after noise reduction preprocessing; and obtain the discharge risk level based on the evaluation index.

[0048] It should be noted that the above modules correspond to the steps in Embodiment 1, and the examples and application scenarios implemented by the above modules and their corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should also be noted that the above modules can be executed in a computer system as part of the system.

[0049] In further embodiments, the following is also provided: An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the method described in Embodiment 1. For brevity, further details are omitted here.

[0050] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0051] A computer-readable storage medium for storing computer instructions that, when executed by a processor, perform the method of Embodiment 1.

[0052] The method in Example 1 can be directly executed by a hardware processor, or it can be executed by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0053] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A transformer discharge detection method with anti-interference function, characterized in that, include: The original signal of multimode discharge of transformer is obtained, and noise reduction preprocessing is performed on the original signal of multimode discharge of transformer by calculating the channel synchronization coefficient and anchoring the synchronization interference signal. Based on the original multi-mode discharge signal of the transformer after noise reduction and preprocessing, initial multi-dimensional features are extracted, and the initial multi-dimensional features are weighted and averaged according to the channel synchronization coefficient to obtain the anti-interference feature vector. A lightweight neural network model is constructed and an interference immune layer is added to obtain a feature selection-interference immune lightweight neural network model. The anti-interference feature vector is input, the channel synchronization coefficient is used as the interference immune factor, and the discharge type identification result is output. Based on the discharge type identification results, an evaluation index is obtained by combining the peak amplitude of the original multi-mode discharge signal of the transformer after noise reduction preprocessing, and the discharge risk level is obtained based on the evaluation index.

2. The transformer discharge detection method with anti-interference function as described in claim 1, characterized in that, The original multi-mode discharge signal of the transformer includes electromagnetic pulse signal, vibration signal and transient voltage signal; the noise reduction preprocessing is specifically as follows: based on the original multi-mode discharge signal of the transformer, the channel synchronization coefficient of the three channels is first calculated, the synchronization interference signal is anchored according to the channel synchronization coefficient, the channel with the weakest interference signal intensity is selected as the reference channel, and the signals of the other two channels are differentially calculated with the reference channel signal to obtain the preprocessed original multi-mode discharge signal of the transformer.

3. The transformer discharge detection method with anti-interference function as described in claim 2, characterized in that, The channel synchronization coefficient is calculated as follows: ; in, The amplitude of the electromagnetic pulse signal. The amplitude of the vibration signal. The amplitude of the transient voltage signal. The average value of the electromagnetic pulse signal amplitude at all times. The average value of the vibration signal amplitude at all times. The average value of the transient voltage signal amplitude over all time points. Total sampling time The sampling time.

4. The transformer discharge detection method with anti-interference function as described in claim 1, characterized in that, The initial multidimensional features are weighted and averaged based on the channel synchronization coefficient, and expressed as follows: ;in, For the first Dimensional fusion features For the first The channel synchronization coefficient corresponding to the dimensional feature. For the first Initial multidimensional features; Arrange all fused features in a fixed order to obtain the anti-interference feature vector.

5. The transformer discharge detection method with anti-interference function as described in claim 1, characterized in that, The feature selection-interference immunity lightweight neural network model sequentially includes an input layer, a convolutional layer, an interference immunity layer, and an output layer. An anti-interference feature vector is input into the input layer, normalized, and then output to the convolutional layer to extract local key features. An activation function is then applied after the convolutional layer, followed by pooling, and the output convolutional feature vector is sent to the interference immunity layer. The channel synchronization coefficient is used as the interference immunity factor to weight and correct the convolutional feature vector, and the interference-immunized feature vector is output to the output layer. The output layer uses a fully connected layer structure to output the discharge type identification result.

6. The transformer discharge detection method with anti-interference function as described in claim 5, characterized in that, The formula for weighted correction of the convolutional feature vector is as follows: ;in, To interfere with the characteristics of post-immunization, The convolutional feature vector This is the channel synchronization coefficient.

7. The transformer discharge detection method with anti-interference function as described in claim 1, characterized in that, The formula for calculating the evaluation index is as follows: ;in, To evaluate the index, This represents the normalized peak amplitude of the original multimode discharge signal from the transformer after noise reduction preprocessing. This is the weighting coefficient for the discharge type.

8. A transformer discharge detection system with anti-interference function, characterized in that, include: The data acquisition and preprocessing module is configured to: acquire the original signal of transformer multimode discharge, calculate the channel synchronization coefficient and anchor the synchronization interference signal, and perform noise reduction preprocessing on the original signal of transformer multimode discharge; The feature extraction module is configured to: extract initial multi-dimensional features based on the original multi-mode discharge signal of the transformer after noise reduction preprocessing; and perform weighted average fusion of the initial multi-dimensional features according to the channel synchronization coefficient to obtain an anti-interference feature vector. The identification and detection module is configured to: construct a lightweight neural network model and add an interference immune layer to obtain a feature screening-interference immune lightweight neural network model, input an anti-interference feature vector, use the channel synchronization coefficient as the interference immune factor, and output the discharge type identification result; The evaluation module is configured to: obtain an evaluation index based on the discharge type identification result and the peak amplitude of the original multi-mode discharge signal of the transformer after noise reduction preprocessing; and obtain the discharge risk level based on the evaluation index.

9. An electronic device, characterized in that, The method includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When the processor executes the computer instructions, it performs the transformer discharge detection method with anti-interference function as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, complete the transformer discharge detection method with anti-interference function as described in any one of claims 1-7.

Images