Valve side sleeve partial discharge identification method and system based on artificial intelligence

Abstract

The invention discloses a valve side sleeve partial discharge identification method and system based on artificial intelligence, and relates to the technical field of power equipment state monitoring and fault diagnosis, and the method comprises the steps: synchronously collecting multi-source signals; pre-processing and feature fusion are carried out based on the multi-source signals, and a fusion feature matrix of partial discharge identification is obtained; and based on the fusion feature matrix of partial discharge identification, performing partial discharge type identification and classification through an artificial intelligence hybrid model. According to the method, the identification robustness of the model to multi-type discharge behaviors in a complex scene is improved, high-quality input with more complete semantic expression and more reasonable structure is provided for a subsequent artificial intelligence model, accurate identification and stable classification of partial discharge types are realized, the probability of false alarm and missing report is effectively reduced, and the accuracy of partial discharge behavior identification is improved. And meanwhile, the system has online learning and model evolution capabilities, so that the system can still keep a high recognition rate in long-term operation, and the dual requirements of an intelligent power grid on online monitoring accuracy and self-adaptability are met.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment condition monitoring and fault diagnosis, and specifically to a method and system for identifying partial discharge of valve side bushings based on artificial intelligence. Background Technique

[0002] With the continuous expansion of the scale of the power system and the continuous improvement of the voltage level, the insulation reliability of large transformers and their key components has gradually become the core technical issue for ensuring the safe operation of the power grid. Among them, the transformer bushing, as an important insulation component connecting the main transformer and the external power grid, undertakes the transmission tasks of high voltage and large current, and its insulation state is directly related to the operation stability of power equipment. In particular, the valve side bushing of the DC converter valve becomes a weak area prone to partial discharge due to its special structural form and operating conditions. Partial discharge, as a typical precursor to the insulation deterioration of electrical equipment, its early detection and accurate identification are of great significance for realizing fault warning and condition maintenance. Currently, the detection methods for partial discharge have gradually evolved from single physical quantity perception means such as traditional pulse current method, ultrasonic detection, and electromagnetic radiation detection to a composite detection system of multi-modal collaborative perception and intelligent identification. Specifically, the ultra-high frequency (UHF) technology can achieve sensitive capture of electromagnetic radiation signals and is suitable for identifying far-field high-frequency signals; the ultrasonic sensing technology has good spatial positioning characteristics and can be used to determine the location of the partial discharge source; while the transient earth voltage (TEV) signal reflects the characteristics of the partial discharge current propagating along the equipment grounding path and is particularly sensitive to near-field partial discharge. Although various detection methods have achieved certain application results in the laboratory and engineering sites, due to the increasingly complex operating environment of power equipment, the types of partial discharge are diverse and the signal forms are non-stationary, the signal-to-noise ratio of a single detection signal is low under the complex interference background, which is prone to problems such as incomplete information and unstable identification results.

[0003] To improve the accuracy and robustness of partial discharge recognition, in recent years, researchers have begun to introduce multi-source signal fusion and artificial intelligence modeling techniques, aiming to break through the limitations of traditional methods in terms of information dimension and spatio-temporal resolution. Typical research paths include: joint criterion methods based on feature fusion, multi-channel time-domain synchronization mechanisms, direction-of-arrival (DOA) positioning algorithms, and end-to-end classification models based on deep learning. Especially in intelligent recognition, convolutional neural networks (CNNs), long short-term memory networks (LSTMs), and their variants have gradually become the mainstream algorithm frameworks in fault diagnosis. Although some academic achievements have attempted to jointly analyze UHF, TEV, and ultrasonic signals and apply deep learning models for recognition, the following problems still generally exist: the signal fusion methods are mostly simple splicing or weighted averaging, without considering the importance differences of different modal signals in spatio-temporal feature expression, and the value of the fused feature information is limited; the design of the preprocessing steps is insufficient, and the noise is not fully removed, resulting in unstable model training performance and weak generalization ability; most deep models adopt a single network structure, making it difficult to simultaneously consider local transient features and long-time series trends, and it is difficult to meet the stable recognition requirements in multi-type partial discharge scenarios; most of the existing methods do not implement the dynamic update and online adaptive capabilities of the model and cannot cope with changes in factors such as the sensing environment, electrical load, and equipment aging status in actual working conditions. Therefore, how to construct a partial discharge identification method with the capabilities of multi-source information collaborative modeling, accurate feature extraction, multi-modal fusion expression, and intelligent evolvable recognition has become a research hotspot and technical bottleneck in the field of high-voltage equipment condition perception. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is that the existing on-line partial discharge monitoring methods have problems such as single signal perception dimension, limited feature extraction quality, single intelligent recognition model structure, and lack of adaptive ability, and how to implement a smart identification scheme for partial discharge types based on synchronous acquisition, differential feature extraction, and dynamic fusion of multi-modal high-frequency signals, constructing an artificial intelligence hybrid model that combines local and global features, and having a self-learning and model optimization mechanism.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: An artificial intelligence-based method for identifying partial discharge in valve-side bushings, including synchronously collecting multi-source signals; performing preprocessing and feature fusion on the multi-source signals to obtain a fusion feature matrix for partial discharge identification; based on the fusion feature matrix for partial discharge identification, performing partial discharge type identification and classification through an artificial intelligence hybrid model; the fusion feature matrix for partial discharge identification includes adjusting the attention weight coefficient of the fusion feature matrix, dynamically evaluating the contribution degree of modal features to discharge type recognition through the attention mechanism, and obtaining a weight-enhanced fusion feature matrix; the artificial intelligence hybrid model includes using a one-dimensional convolutional neural network to extract local features from the fusion feature matrix, extracting local spatial and short-term temporal information features from the fusion feature matrix through one-dimensional convolutional kernels to obtain local feature representations, extracting long-term dependence global temporal features from the local feature representations to form global temporal feature representations, fusing the local feature representations and global temporal feature representations in a skip connection manner to form a hybrid feature vector containing both short-term local and long-term global information, and inputting the hybrid feature vector into a classification layer to output the partial discharge type recognition and classification results through the Softmax function.

[0007] As a preferred embodiment of the artificial intelligence-based method for identifying partial discharge in valve-side bushings according to the present invention, wherein: the synchronously collecting multi-source signals includes collecting electromagnetic wave signals through a UHF sensor, collecting ultrasonic wave signals through an ultrasonic sensor, collecting transient earth voltage signals through a transient earth voltage sensor, and assigning a consistent time tag to each group of data during the sampling process to synchronously collect data.

[0008] As a preferred embodiment of the artificial intelligence-based method for identifying partial discharge in valve-side bushings according to the present invention, wherein: the preprocessing includes denoising the electromagnetic wave signals using the wavelet packet decomposition method, extracting the time-frequency domain characteristic parameters of the electromagnetic wave signals, performing spatial positioning on the ultrasonic wave signals based on the direction of arrival algorithm, and extracting the pulse amplitude and change rate characteristics of the transient earth voltage signals.

[0009] The feature fusion includes aligning the time-frequency domain characteristic parameters of the electromagnetic wave signals and the pulse amplitude and change rate characteristics of the transient earth voltage signals to obtain a feature vector.

[0010] As a preferred embodiment of the artificial intelligence-based method for identifying partial discharge in valve-side bushings according to the present invention, wherein: the fusion feature matrix for partial discharge identification includes calculating the correlation weight coefficient between feature vectors based on the attention mechanism, performing weighted fusion on the feature vectors, and generating a weighted fusion feature matrix.

[0011] As a preferred solution of the method for identifying partial discharge of valve-side bushings based on artificial intelligence according to the present invention, wherein: the artificial intelligence hybrid model includes using a residual network to extract spatial features from the fused feature matrix to obtain a multi-scale spatial feature map, using a Transformer encoder to extract temporal features from the fused feature matrix to obtain a temporal feature representation with long-term dependence, and performing feature concatenation on the spatial feature map and the temporal feature representation to form a comprehensive feature vector integrating spatial and temporal information.

[0012] As a preferred solution of the method for identifying partial discharge of valve-side bushings based on artificial intelligence according to the present invention, wherein: the dynamic adjustment of the alarm threshold according to the discharge type and the current load rate includes obtaining the real-time operating load rate information of the current device, calling a preset alarm threshold based on the discharge type, matching an adjustment factor according to the real-time operating load rate information, using the adjustment factor to weight and correct the reference threshold in the alarm threshold, dynamically adjusting the alarm threshold, extracting the fused feature parameters in the fused feature vector, comparing the fused feature parameters with the alarm threshold, and determining whether to trigger an alarm.

[0013] As a preferred solution of the method for identifying partial discharge of valve-side bushings based on artificial intelligence according to the present invention, wherein: the adaptive optimization of the artificial intelligence hybrid model includes collecting the fused feature matrix and the actual discharge type labels in real time, updating the historical data set, retraining the updated historical data set, optimizing the network weights and bias parameters in the artificial intelligence hybrid model using the gradient descent method, dynamically updating the model weight parameters based on the latest training results, and adjusting the probability threshold of the Softmax classifier in real time. Through continuous iterative updates, the parameters of the artificial intelligence hybrid model are adaptively optimized.

[0014] Another object of the present invention is to provide a system for identifying partial discharge of valve-side bushings based on artificial intelligence, which can realize synchronous acquisition, feature fusion and intelligent recognition and early warning of multi-modal high-frequency signals, and solves the problems of difficult accurate identification of partial discharge signal types, strong noise interference and insufficient spatial positioning accuracy in the prior art.

[0015] As a preferred solution of the artificial intelligence-based partial discharge identification system for valve-side bushings of the present invention, it includes a signal acquisition module, a feature processing module, and an intelligent identification module; the signal acquisition module includes a multi-source acquisition unit and a time-domain synchronization unit. The multi-source acquisition unit is used to synchronously acquire UHF, ultrasonic, and transient earth voltage signals, and the time-domain synchronization unit is used to add a unified time stamp to the acquired signals to achieve time-domain alignment of multi-source signals; the feature processing module includes a pre-noise reduction unit and a feature fusion unit. The pre-noise reduction unit is used to remove noise from the multi-modal original signals and extract effective features, and the feature fusion unit is used to weight and fuse the multi-modal features through an attention mechanism to form a fused feature matrix; the intelligent identification module includes a hybrid modeling unit and an adaptive optimization unit. The hybrid modeling unit is used to extract features and classify the fused feature matrix using an artificial intelligence hybrid model, and the adaptive optimization unit is used to optimize the model parameters in real time according to on-site feedback data.

[0016] A computer device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it realizes the steps of the artificial intelligence-based partial discharge identification method for valve-side bushings.

[0017] A computer-readable storage medium stores a computer program thereon. When the computer program is executed by a processor, it realizes the steps of the artificial intelligence-based partial discharge identification method for valve-side bushings.

[0018] The beneficial effects of the present invention: By setting a unified time reference and synchronously acquiring multi-source signals, parallel acquisition of electromagnetic wave signals, ultrasonic signals, and transient earth voltage signals in a unified time window is achieved, thus establishing a multi-modal and high-time-consistency raw data basis. By accurately extracting partial discharge features and constructing a fused feature matrix in a weighted fusion manner, the identification robustness of the model for multi-type discharge behaviors in complex scenarios is improved, providing high-quality input with more complete semantic expression and more reasonable structure for the subsequent artificial intelligence model, thereby enhancing the recognition ability of the classifier for discharge patterns under weak and multi-source interference conditions. By constructing an artificial intelligence hybrid architecture with spatio-temporal collaborative modeling capabilities, accurate identification and stable classification of partial discharge types are achieved, effectively reducing the probability of false alarms and missed alarms, and at the same time having the ability of online learning and model evolution, enabling the system to still maintain a high recognition rate during long-term operation and meeting the dual requirements of the smart grid for the accuracy and self-adaptability of online monitoring. Description of the Drawings

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0020] Figure 1 It is the overall flowchart of a method for identifying partial discharge of valve-side bushings based on artificial intelligence provided for the first embodiment of the present invention. Specific embodiments

[0021] To make the above objects, features, and advantages of the present invention more obvious and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings of the specification. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. 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 scope of protection of the present invention.

[0022] Embodiment 1, referring to Figure 1 , which is an embodiment of the present invention, provides a method for identifying partial discharge of valve-side bushings based on artificial intelligence, including: S1: Synchronously collect multi-source signals.

[0023] Further, synchronously collecting multi-source signals includes collecting electromagnetic wave signals through ultra-high frequency sensors, collecting ultrasonic signals through ultrasonic sensors, collecting transient ground voltage signals through transient ground voltage sensors, and assigning consistent time tags to each group of data during the sampling process to synchronously collect data.

[0024] It should be noted that the electromagnetic radiation signals generated in the frequency band of 300 MHz to 3 GHz during the partial discharge process of the valve side bushing are collected by the ultra-high frequency sensor to reflect the characteristics of the far-field high-frequency electromagnetic energy distribution during the discharge process; the acoustic emission signals in the frequency band of 20 kHz to 200 kHz during the partial discharge process are collected by the ultrasonic sensor to capture the spatial propagation path and the change of sound pressure amplitude of the discharge source in the bushing structure; the transient current pulse signals generated in the bushing grounding wire during the partial discharge are collected by the transient earth voltage sensor to identify the coupling effect and pulse response form of the discharge along the grounding loop. During the sampling process, to ensure the alignment consistency of multi-modal signals in the time dimension, a unified clock reference is configured for all data sampling channels, and a consistent time tag is assigned to each set of collected data to achieve synchronous acquisition and time-domain registration of multi-source signals. Different types of signal sources have their own advantages in response mechanisms, propagation paths, signal characteristics, and anti-interference capabilities. The UHF (ultra-high frequency) signal can capture the change of spatial radiation energy with high sensitivity, the ultrasonic signal has strong discharge sound source localization ability, and the TEV (transient earth voltage) signal can reveal the conduction trajectory of the discharge current in the metal grounding structure. Through this multi-source collaborative acquisition mechanism, multi-dimensional and multi-angle response information is obtained during the occurrence of the same discharge event, constructing a raw data input with strong complementarity and unified time sequence, laying a solid data foundation for subsequent feature extraction, fusion, and artificial intelligence modeling, thereby effectively enhancing the recognition robustness of the system to partial discharge signals and the accuracy of discharge type discrimination under complex working conditions.

[0025] S2: Based on the multi-source signals, perform preprocessing and feature fusion to obtain the fusion feature matrix for partial discharge identification.

[0026] Furthermore, the preprocessing includes denoising the electromagnetic wave signals using the wavelet packet decomposition method, extracting the time-frequency domain characteristic parameters of the electromagnetic wave signals, performing spatial positioning on the ultrasonic signals based on the direction of arrival algorithm, and extracting the pulse amplitude and change rate characteristics of the transient earth voltage signals.

[0027] The feature fusion includes aligning the time-frequency domain characteristic parameters of the electromagnetic wave signals, the pulse amplitude and change rate characteristics of the transient earth voltage signals to obtain the feature vectors.

[0028] It should be noted that the collected electromagnetic wave signals are denoised by the wavelet packet decomposition method. The sym6 wavelet basis function, which has good response ability to mutation points, is selected as the decomposition basis. The decomposition layer is set to 6 layers. The high-frequency transient characteristics contained in the partial discharge signals are extracted by the hierarchical frequency band expansion method, and the effective high-frequency pulse components are retained while suppressing the background low-frequency interference, so as to improve the feature fidelity of the electromagnetic wave signals in the complex interference environment. Further, the time-frequency domain characteristic parameters of the electromagnetic wave signals, including pulse peak value, rise time, and spectral centroid, etc., are extracted from the effective frequency bands after wavelet decomposition and used as the input for the subsequent recognition of discharge type patterns. For the collected ultrasonic signals, the sound source spatial positioning is carried out by using the direction of arrival algorithm. For the transient earth voltage signals, the pulse amplitude and change rate characteristics are extracted, which respectively represent the absolute intensity of the energy released by the partial discharge event and the steepness of the signal edge, so as to enhance the perception ability of low-amplitude and fast-changing discharge behaviors.

[0029] Feature fusion includes unifying the dimensional alignment and feature matching of the key feature vectors extracted from the multi-source sensing signals after preprocessing. This includes pairwise matching of the time-frequency domain statistical characteristic parameters of the electromagnetic wave signals (such as peak value, rise time, spectral centroid, etc.), the pulse amplitude and change rate characteristics of the transient earth voltage signals, etc. to construct a fused feature vector group; for the information redundancy and feature importance differences between different modal signals, an attention weighting mechanism is introduced to adaptively adjust the expression weights of each feature in the fusion process; finally, a fused feature vector with complete expression, unified dimension, and high discrimination for partial discharge types is formed as the input for the subsequent artificial intelligence hybrid model. The time-frequency characteristics of the electromagnetic wave signals reflect the energy release mode of the partial discharge behavior in the high-frequency electromagnetic domain and are suitable for capturing corona and far-field discharge characteristics; while the pulse amplitude and edge change rate of the transient earth voltage signals can better reflect the transient impact characteristics of the near-field discharge current. By fusing the above multi-modal heterogeneous information and jointly modeling at the feature level, the system can obtain higher information integrity and robustness, thus effectively improving the accuracy and adaptability of partial discharge type recognition in complex interference environments.

[0030] The direction of arrival algorithm includes determining the delay time of the ultrasonic wave reaching different sensor elements according to the time difference of the ultrasonic wave signals received by the sensor elements, calculating the phase difference of the received signals of each sensor element based on the delay time, constructing a spatial steering vector, and using the spatial spectrum estimation method to search for the spatial angle when the matching degree between the steering vector and the measured data reaches the maximum, so as to determine the direction of arrival of the ultrasonic wave signal. The determination of the delay time of the ultrasonic wave reaching different sensor elements is expressed as: ; Among them, represents the The delay time when an ultrasonic sensor element receives a partial discharge acoustic emission signal Indicates the sensor serial number in the sensor array Indicates the center spacing between adjacent sensor elements Indicates the incident angle of the acoustic signal relative to the normal direction of the sensor array Indicates the propagation speed of ultrasonic waves in the medium

[0031] The phase difference of the received signals of each sensor element calculated based on the delay time is expressed as: ; Among them, Indicates the phase difference between the signal received by the th sensor element and the first element, Indicates the frequency of the ultrasonic signal generated by partial discharge

[0032] The spatial steering vector is constructed and expressed as: ; Among them, Indicates the steering vector, Indicates the phase delay term of the th sensor element, Indicates the phase response of the th element in the sensor array to the signal in the sound source direction

[0033] Determine the direction of arrival of the ultrasonic signal and express it as: ; Among them, Indicates the direction of arrival angle, Indicates the response intensity of the array to different incident directions

[0034] The sum of the steering vectors and the search for the maximum value of the spatial spectrum constitute the core of the direction-of-arrival estimation algorithm. Through the data linkage with the multi-channel ultrasonic sensors, it can achieve high-precision inversion of the spatial position of the partial discharge source, provide spatial constraint input for the subsequent fusion feature construction and artificial intelligence model, and greatly improve the interpretability and discrimination performance of the system for partial discharge types and source points.

[0035] Furthermore, the fusion feature matrix for partial discharge identification includes calculating the correlation weight coefficient between feature vectors based on the attention mechanism, performing weighted fusion on the feature vectors, and generating a weighted fusion feature matrix.

[0036] It should also be noted that calculating the correlation weight coefficient between feature vectors based on the attention mechanism is expressed as: Construct an electromagnetic wave feature vector , the modal features are uniformly transformed to form an input feature matrix , in the attention mechanism: ; Among them, represents the feature vector matrix of the ultra-high frequency electromagnetic wave signal, represents the feature vector matrix of the ultrasonic wave signal, represents the feature vector matrix of the transient earth voltage signal, is the number of features, the feature dimension, , , respectively represent the query, key, and value vector representations, , , are trainable projection matrices.

[0037] The calculation of the attention weight matrix is expressed as: ; Among them, represents the attention weight matrix, represents normalizing each row in the matrix to convert it into a probability distribution for controlling the weight distribution of each feature dimension, represents the transpose of the key matrix, represents the dimension of the key vector.

[0038] The generation of the fused feature matrix after weighted fusion is expressed as: ; Among them, represents the fused feature matrix, which is used as the input of the subsequent artificial intelligence hybrid model, represents the value matrix, that is, the feature vector matrix before fusion.

[0039] S3: Based on the fused feature matrix for partial discharge identification, the type identification and classification of partial discharges are carried out through an artificial intelligence hybrid model.

[0040] Furthermore, the artificial intelligence hybrid model includes using a residual network to extract spatial features from the fused feature matrix to obtain multi-scale spatial feature maps, using a Transformer encoder to extract temporal features from the fused feature matrix to obtain temporal feature representations with long-term dependencies, and performing feature concatenation on the spatial feature maps and temporal feature representations to form a comprehensive feature vector that fuses spatial and temporal information.

[0041] It should be noted that the artificial intelligence hybrid model is expressed as: ; Among them, represents the comprehensive feature vector after fusing spatial and temporal information, is the spatial feature map representation, is the temporal feature representation, represents concatenation (cascading) along the feature dimension to obtain the joint feature vector of spatial and temporal information.

[0042] Based on the fused feature matrix, a joint modeling architecture is constructed. The residual neural network (Res Net) is used to extract the multi-scale spatial feature representation in the fused feature matrix to obtain the spatial feature map representing the distribution relationship of different modal partial discharge features in the spatial dimension; the Transformer encoder is used to globally model the fused feature matrix, and the long-term dependence relationship of each feature vector in the time series is extracted through the multi-head attention mechanism to construct the long-term dependence temporal feature representation representing the change trend of partial discharge behavior. The spatial feature map and the temporal feature representation are cascaded at the feature level through the skip connection structure to form a joint expression vector that contains both local spatial features and global temporal structures; the artificial intelligence hybrid model combines the deep convolutional expression ability of Res Net in spatial modeling and the long-distance dependence capture ability of Transformer in sequence modeling, which can effectively solve the modeling balance problem between the local sensitivity and global discriminability of multi-modal fused features, and improve the stability and accuracy of partial discharge type identification under abnormal conditions such as complex background noise and modal feature loss.

[0043] Furthermore, the partial discharge type identification and classification include, according to the comprehensive feature vector, calculating the probability distribution of the comprehensive feature vector corresponding to each discharge type through the Softmax classifier, sorting the probabilities of each discharge type using the probability threshold discrimination method, determining the partial discharge type and outputting the discharge type label, and performing confidence analysis on the output discharge type label. When the confidence is lower than the preset threshold, the historical feature data is called to recheck and adjust the parameters of the artificial intelligence hybrid model, and the artificial intelligence hybrid model is adaptively optimized.

[0044] It should be noted that calculating the probability distribution of the comprehensive feature vector corresponding to each discharge type is expressed as: ; where, represents the output discharge type classification probability vector, represents the multi-class normalization function, outputting the discharge type probability distribution, represents the weight matrix of the classifier, represents the bias vector of the classifier.

[0045] The fused comprehensive feature vector is input into the Softmax classifier to calculate the probability distribution of various partial discharge types (such as corona discharge, surface discharge, internal discharge, etc.), and the confidence probability value corresponding to each discharge type is output. The Softmax function maps the fused feature vector in the multi-class space to make the output values meet the normalization condition, forming the probability prediction results for all predefined discharge types. Based on the probability output, the maximum probability criterion is used to discriminate and rank various partial discharge types to determine the most likely corresponding discharge type of the current sample, and then the discharge type label is output. To improve the reliability and recognition accuracy of the model, the system further performs confidence analysis on the output discharge type label: when the maximum probability value output by the classifier is lower than the set confidence threshold, the system triggers an adaptive feedback mechanism, automatically calls the historical feature vector library, performs similarity matching and label re-estimation on the current sample and the existing samples, corrects the decision boundary of the model through the resampling mechanism, and incorporates the re-verified samples and historical samples into the joint training set to fine-tune the parameters of the artificial intelligence hybrid model and update the network weight matrix, so as to achieve the adaptive optimization and dynamic evolution of the model structure in the actual operating environment. This mechanism significantly improves the robustness and adaptability of the model under complex electromagnetic environments, multi-source modality deficiencies, and rare discharge type conditions, ensuring the long-term stability and high credibility of the partial discharge type classification results.

[0046] Furthermore, the adaptive optimization of the artificial intelligence hybrid model includes: collecting the fused feature matrix and the actual discharge type label in real time, updating the historical data set, retraining the updated historical data set, optimizing the network weights and bias parameters in the artificial intelligence hybrid model using the gradient descent method, dynamically updating the model weight parameters based on the latest training results, and adjusting the probability threshold of the Softmax classifier in real time. Through continuous iterative updates, the parameters of the artificial intelligence hybrid model are adaptively optimized.

[0047] It should be noted that during operation, the fused feature matrix and the actual output partial discharge type labels are collected in real time, a new sample data set is continuously constructed, and the updated samples are merged with the existing historical data to form a dynamic historical sample library. For the historical sample library, the system triggers incremental training of the artificial intelligence hybrid model, and retrains the neural network structure in the model using an optimization algorithm based on gradient descent, including updating the weight parameters and bias parameters in the convolutional layer and the Transformer encoder layer, so that the model can continuously adapt to the changes in the discharge feature distribution under the latest working conditions. During the training process, the system dynamically monitors the training process based on the loss function and the classification performance index (accuracy), judges the convergence state of the model in real time, and adjusts the learning rate of the network parameters, the regularization factor, etc. accordingly to improve the training efficiency and stability. At the same time, the Softmax classifier at the output end is updated online. For the fuzzy classification boundary or new partial discharge type in the new samples, the probability threshold distribution of each category is re-estimated to ensure the reliability and discrimination of the model classification output. By continuously iteratively fine-tuning the model and enhancing the structure, the artificial intelligence hybrid model has the ability of self-evolution and self-adaptation, and can cope with the feature migration problems caused by factors such as equipment aging, noise interference, and environmental disturbances during long-term operation, so as to maintain the high stability and high generalization performance of the partial discharge identification system, and effectively meet the demand for continuous optimization of the performance of the online monitoring system in the complex power grid operation environment.

[0048] Embodiment 2 is an embodiment of the present invention, which provides an artificial intelligence-based valve side bushing partial discharge identification system, including a signal acquisition module, a feature processing module, and an intelligent identification module.

[0049] Among them: The signal acquisition module includes a multi-source acquisition unit and a time-domain synchronization unit. The multi-source acquisition unit is used to synchronously acquire UHF, ultrasonic, and transient earth voltage signals, and the time-domain synchronization unit is used to add a unified time stamp to the acquired signals to achieve time-domain alignment of multi-source signals.

[0050] It should also be noted that the multi-source acquisition unit transmits the acquired multi-modal signals to the time-domain synchronization unit, and the synchronized signals output by the time-domain synchronization unit are transmitted to the pre-noise reduction unit.

[0051] The feature processing module includes a pre-noise reduction unit and a feature fusion unit. The pre-noise reduction unit is used to remove the noise in the multi-modal original signals and extract effective features, and the feature fusion unit is used to weight-fuse the multi-modal features through an attention mechanism to form a fused feature matrix.

[0052] It should also be noted that the denoised features output by the pre-noise reduction unit are transmitted to the feature fusion unit, and the fused feature matrix output by the feature fusion unit is transmitted to the hybrid modeling unit.

[0053] The intelligent identification module includes a hybrid modeling unit and an adaptive optimization unit. The hybrid modeling unit is used to extract and classify features from the fusion feature matrix using an artificial intelligence hybrid model, and the adaptive optimization unit is used to optimize the model parameters in real time according to on-site feedback data.

[0054] It should also be noted that the classification results output by the hybrid modeling unit and the feedback information are transmitted to the adaptive optimization unit.

Claims

1. A method for identifying partial discharge in valve-side bushings based on artificial intelligence, characterized in that, Including: Synchronously collect multi-source signals; Based on the multi-source signals, perform preprocessing and feature fusion to obtain a fusion feature matrix for partial discharge identification; Based on the fusion feature matrix for partial discharge identification, perform partial discharge type identification and classification through an artificial intelligence hybrid model; The fusion feature matrix for partial discharge identification includes adjusting the attention weight coefficients of the fusion feature matrix, dynamically evaluating the contribution degree of modal features to discharge type recognition through the attention mechanism, and obtaining a fusion feature matrix with enhanced weights; The artificial intelligence hybrid model includes using a one-dimensional convolutional neural network to extract local features from the fusion feature matrix, extracting local spatial and short-term temporal information features from the fusion feature matrix through one-dimensional convolutional kernels to obtain local feature representations, extracting long-term dependent global temporal features from the local feature representations to form global temporal feature representations, fusing the local feature representations and global temporal feature representations through skip connection methods to form a hybrid feature vector containing both short-term local and long-term global information, inputting the hybrid feature vector into the classification layer, and outputting the partial discharge type recognition and classification results through the Softmax function.

2. The method for identifying partial discharge of valve-side bushing based on artificial intelligence according to claim 1, wherein: The synchronous collection of multi-source signals includes collecting electromagnetic wave signals through ultra-high frequency sensors, collecting ultrasonic signals through ultrasonic sensors, collecting transient earth voltage signals through transient earth voltage sensors, and assigning consistent time tags to each group of data during the sampling process to synchronously collect data.

3. The method for identifying partial discharge in the valve side bushing based on artificial intelligence according to claim 2, wherein: The preprocessing includes denoising the electromagnetic wave signals using the wavelet packet decomposition method, extracting the time-frequency domain characteristic parameters of the electromagnetic wave signals, performing spatial positioning on the ultrasonic signals based on the direction of arrival algorithm, and extracting the pulse amplitude and change rate characteristics of the transient earth voltage signals; Feature fusion includes: Align the time-frequency domain characteristic parameters of the electromagnetic wave signals and the pulse amplitude and change rate characteristics of the transient earth voltage signals to obtain feature vectors.

4. The method for identifying partial discharge of valve side bushing based on artificial intelligence according to claim 3, characterized in that: The fusion feature matrix for partial discharge identification includes calculating the correlation weight coefficients between feature vectors based on the attention mechanism, and performing weighted fusion on the feature vectors to generate a weighted fusion feature matrix.

5. The method for identifying partial discharge of valve side bushing based on artificial intelligence according to claim 4, characterized in that: The artificial intelligence hybrid model includes using a residual network to extract spatial features from the fusion feature matrix to obtain multi-scale spatial feature maps, using a Transformer encoder to extract temporal features from the fusion feature matrix to obtain long-term dependent temporal feature representations, and performing feature concatenation on the spatial feature maps and temporal feature representations to form a comprehensive feature vector integrating spatial and temporal information.

6. The method for identifying partial discharge in valve side bushings based on artificial intelligence according to claim 5, characterized in that: The partial discharge type identification and classification includes, according to the comprehensive feature vector, calculating the probability distribution of the comprehensive feature vector corresponding to various discharge types through a Softmax classifier, sorting the probabilities of various discharge types using the probability threshold discrimination method, determining the partial discharge type and outputting the discharge type label, performing confidence analysis on the output discharge type label, when the confidence is lower than the preset threshold, calling historical feature data to recheck and adjust the parameters of the artificial intelligence hybrid model, and adaptively optimizing the artificial intelligence hybrid model.

7. The method for identifying partial discharge of valve side bushing based on artificial intelligence according to claim 6, wherein: The adaptive optimization of the artificial intelligence hybrid model includes: collecting the fusion feature matrix and the actual discharge type labels in real time, updating the historical data set, retraining the updated historical data set, optimizing the network weights and bias parameters in the artificial intelligence hybrid model using the gradient descent method, dynamically updating the model weight parameters based on the latest training results, and adjusting the probability threshold of the Softmax classifier in real time. Through continuous iterative updates, the parameters of the artificial intelligence hybrid model are adaptively optimized.

8. A system adopting the artificial intelligence-based local discharge identification method for valve side bushings according to any one of claims 1 to 7, characterized in that: It includes a signal acquisition module, a feature processing module, and an intelligent identification module; The signal acquisition module includes a multi-source acquisition unit and a time-domain synchronization unit. The multi-source acquisition unit is used to synchronously acquire UHF, ultrasonic, and transient earth voltage signals. The time-domain synchronization unit is used to add a unified time stamp to the acquired signals to achieve time-domain alignment of multi-source signals; The feature processing module includes a pre-noise reduction unit and a feature fusion unit. The pre-noise reduction unit is used to remove the noise in the multi-modal original signals and extract effective features. The feature fusion unit is used to weight-fuse the multi-modal features through an attention mechanism to form a fusion feature matrix; The intelligent identification module includes a hybrid modeling unit and an adaptive optimization unit. The hybrid modeling unit is used to extract features and classify the fusion feature matrix using the artificial intelligence hybrid model. The adaptive optimization unit is used to optimize the model parameters in real time according to the on-site feedback data.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the artificial intelligence-based valve-side bushing partial discharge identification method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the artificial intelligence-based valve-side bushing partial discharge identification method according to any one of claims 1 to 7.

Images