Partial discharge intelligent identification method and system based on collaborative reasoning

Abstract

The present invention provides a collaborative reasoning-based intelligent identification method and system for partial discharge (PD). This system belongs to the field of partial discharge detection. By constructing a multi-scale feature information graph and using a sparse spectrum partitioning algorithm, the semantic structure of high-dimensional PD data is explicitly deconstructed, separating feature sub-channels strongly correlated with the discharge mechanism and effectively suppressing semantic interference in field data. Furthermore, a channel-structure-induced fuzzy modeler is used to independently learn local rule sets. Combined with a dynamic fusion mechanism driven by cross-channel prediction consistency, this method significantly improves tolerance to voltage phase loss, background noise disturbances, and device heterogeneity. Compared to traditional neural fuzzy models, this method achieves non-exclusive fuzzy classification of unknown discharge patterns while maintaining interpretability. This solves the problem of model failure caused by data distribution offsets during field deployment, significantly reduces recognition error rates, and provides a highly robust solution for intelligent diagnosis of power equipment.

Description

Technical Field

[0001] The present invention relates to the field of partial discharge detection, and in particular to a partial discharge intelligent identification method and system based on collaborative reasoning. Background Art

[0002] Partial discharge (PD) identification is a core method for monitoring the insulation condition of power equipment. Traditional methods rely on clean voltage-discharge synchronization signals in laboratory settings and employ neural networks or fuzzy systems for pattern classification. Such techniques perform well in controlled environments, such as GIS partial discharge identification based on fractal features (such as box dimension and information dimension) (see case patent CN 104155585 A). However, field conditions present the challenge of multi-source heterogeneous data: differences in sensor models, varying installation methods, diverse cable types, and operating voltage fluctuations lead to significant shifts in the data distribution structure. More critically, field measurements often lack voltage phase labels, which serve as a key input feature in laboratory models, resulting in a significant reduction in the generalization ability of existing models when deployed in the field.

[0003] Current PD identification methods suffer from four limitations: First, high-dimensional measurement data contains a large number of semantically interfering sub-channels, but existing models lack a dynamic channel-level structural modeling mechanism, making it impossible to effectively separate interference. Second, they fail to consider the differences in the coupling structures of different discharge types (such as corona discharge and suspended solids discharge) in the feature dimension, resulting in insufficient feature expression capabilities. Third, traditional fuzzy inference rules lack synergy with feature groups, limiting rule generalization and making it difficult to adapt to complex field noise backgrounds. Fourth, existing frameworks cannot uniformly address inherent field challenges such as missing voltage phases and strong background noise. For example, a patented case relies on power frequency period segmentation to extract half-cycle fractal features, but this solution fails when phase labels are unavailable in the field. These issues lead to increased recognition error rates and low tolerance for unknown patterns, severely restricting engineering practicality.

[0004] Therefore, there is an urgent need to build a new recognition framework that takes into account both structural deconstruction capabilities and fuzzy reasoning synergy, so as to solve special field challenges such as semantic interference separation in high-dimensional heterogeneous data, cross-device data distribution offset, and voltage phase loss, and improve the model's portability and robustness. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for intelligent identification of partial discharge based on collaborative reasoning to solve the problems existing in the above-mentioned prior art.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] The present invention provides a method for intelligent identification of partial discharge based on collaborative reasoning, comprising the following steps:

[0008] S1. Construct a multi-scale feature information map of partial discharge measurement data and extract high-order dependencies between dimensions;

[0009] S2. Divide the high-dimensional feature space into multiple structurally consistent feature sub-channels using a structural decomposition algorithm, where each sub-channel corresponds to a local feature cluster related to the discharge mechanism;

[0010] S3. Build a fuzzy inference model independently for each characteristic sub-channel to generate a channel-level discharge pattern sub-classifier;

[0011] S4. Dynamically adjust the fusion strategy based on prediction confidence and perform structural consistency-driven fusion of multiple sub-classifier outputs;

[0012] S5. Output the final partial discharge type identification result.

[0013] Preferably, in step S1, the multi-scale feature information graph uses at least one of mutual information, GraphLasso algorithm, kernel correlation coefficient or Copula function to construct the dependency relationship between dimensions.

[0014] Preferably, in step S2, the structural decomposition algorithm is a sparse spectral partitioning algorithm, a hierarchical Louvain clustering algorithm or an adaptive multi-channel clustering algorithm.

[0015] Preferably, in step S3, the fuzzy reasoning model is constructed by automatically learning a fuzzy rule set and a membership function structure.

[0016] Preferably, in step S3, the fuzzy inference model can be replaced by a sparse Bayesian regression model or an integrated fuzzy network model.

[0017] Preferably, in step S4, the dynamically adjusting the fusion strategy includes:

[0018] Calculate the prediction confidence of each sub-classifier;

[0019] Weighted fusion output according to confidence weight;

[0020] Implement a fault-tolerant mechanism for inconsistencies between feature channels.

[0021] Preferably, in step S4, the fusion mechanism can be replaced by a multi-task attention allocation model or a confidence regression network optimization weighted strategy.

[0022] Preferably, in step S5, the recognition result supports non-exclusive fuzzy classification of unknown or borderline discharge patterns.

[0023] The present invention also provides a partial discharge intelligent identification system based on collaborative reasoning, comprising:

[0024] Feature map construction module, used to build multi-scale feature information maps of partial discharge measurement data and extract high-order dependencies between dimensions;

[0025] The channel partitioning module is used to divide the high-dimensional feature space into multiple structurally consistent feature sub-channels through a structural decomposition algorithm. Each sub-channel corresponds to a local feature group related to the discharge mechanism.

[0026] A fuzzy inference module is used to independently construct a fuzzy inference model for each feature sub-channel and generate a channel-level discharge pattern sub-classifier;

[0027] The dynamic fusion module is used to dynamically adjust the fusion strategy based on the prediction confidence and perform structural consistency-driven fusion of multiple sub-classifier outputs.

[0028] Preferably, the dynamic fusion module has a built-in cross-channel prediction consistency verification mechanism, which triggers a confidence re-weighting process for conflicting outputs.

[0029] Compared with the prior art, the present invention has achieved the following beneficial technical effects:

[0030] The present invention provides a collaborative reasoning-based intelligent identification method and system for partial discharge. This method explicitly deconstructs the semantic structure of high-dimensional PD data through the construction of a multi-scale feature information graph and a sparse spectrum partitioning algorithm, separating feature sub-channels strongly correlated with the discharge mechanism and effectively suppressing semantic interference in field data. Furthermore, a channel-structure-induced fuzzy modeler is used to independently learn local rule sets, combined with a dynamic fusion mechanism driven by cross-channel prediction consistency. This significantly improves tolerance to voltage phase loss, background noise disturbances, and device heterogeneity. Compared to traditional neural fuzzy models, this method achieves non-exclusive fuzzy classification of unknown discharge patterns while maintaining interpretability. This addresses the problem of model failure caused by data distribution offsets in field deployments, significantly reduces recognition error rates, and provides a highly robust solution for intelligent diagnosis of power equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0032] Figure 1 This is a flow chart of the partial discharge intelligent identification method based on collaborative reasoning provided by the present invention. DETAILED DESCRIPTION

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0034] The purpose of the present invention is to provide a method and system for intelligent identification of partial discharge based on collaborative reasoning to solve the problems existing in the prior art.

[0035] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0036] Example 1

[0037] This embodiment provides a method for intelligent identification of partial discharge based on collaborative reasoning, comprising the following steps:

[0038] S1. Construct a multi-scale feature information graph of partial discharge measurement data, extract high-order dependencies between dimensions, and build a feature association network;

[0039] First, construct the mutual information matrix :

[0040] ;

[0041] Secondly, the Laplace matrix is ​​used to calculate: ;

[0042] S2. Divide the high-dimensional feature space into multiple structurally consistent feature sub-channels using a structural decomposition algorithm, where each sub-channel corresponds to a local feature cluster related to the discharge mechanism;

[0043] First, yes Perform spectral decomposition and take the front Eigenvectors construct matrices ;

[0044] Secondly, Cluster each row (corresponding to each feature) in Semantic channels: ; Output: channel partition set ;

[0045] S3. Build a fuzzy inference model independently for each characteristic sub-channel, automatically learn the fuzzy rule set and membership function structure, and generate a channel-level discharge pattern sub-classifier;

[0046] For each channel : First, subspace feature extraction: ;

[0047] Second, construct each category The TSK fuzzy model has Rules:

[0048] No. Gaussian membership function of the rules: ;

[0049] Activation Strength: ; Further, each category The output is: ;

[0050] Furthermore, the channel pair The output of is a weighted aggregation: ;

[0051] Finally, all category outputs form a vector: ;

[0052] S4. Dynamically adjust the fusion strategy based on prediction confidence and perform structural consistency-driven fusion of multiple sub-classifier outputs;

[0053] First, the channel model prediction difference is calculated: ;

[0054] Second, weight distribution:

[0055] like (predicting large conflict), using contrast weighting:; ;

[0056] Otherwise, use channel confidence weighting: ;

[0057] Furthermore, multi-output fusion: ;

[0058] S5. Output the final partial discharge type identification result;

[0059] First, the final output is a multidimensional category vector: ;

[0060] in, is the number of categories;

[0061] Secondly, hard classification output (main output path): category decision is made directly through the maximum value: ;

[0062] This output method is suitable for standard single-label classification tasks and can be combined with softmax for probabilistic interpretation: ;

[0063] Furthermore, fuzzy label interpretation is optional (for boundary sample analysis and non-exclusive reasoning): To enhance the model's ability to explain sample uncertainty and fuzzy boundaries between classes, the prediction output can be and class mean Compare and construct a "fuzzy response" function: ;

[0064] in, , represents the sample pair class "Fuzzy response strength" or "similarity"; To control the response width (ambiguity); is the normalization factor.

[0065] As an implementation method, in step S1, the multi-scale feature information graph uses mutual information. Of course, in addition to mutual information, at least one of the GraphLasso algorithm, kernel correlation coefficient, or Copula function can also be used to construct the dependency relationship between dimensions.

[0066] As an implementation manner, in step S2, the structural decomposition algorithm is a sparse spectrum partitioning algorithm. Of course, it can also be replaced by methods such as hierarchical Louvain clustering and adaptive multi-channel clustering.

[0067] As an implementation method, in step S3, the fuzzy inference model can be replaced by a sparse Bayesian regression model or an integrated fuzzy network model.

[0068] As an implementation method, in step S4, dynamically adjusting the fusion strategy includes:

[0069] Calculate the prediction confidence of each sub-classifier;

[0070] Weighted fusion output according to confidence weight;

[0071] Implement a fault-tolerant mechanism for inconsistencies between feature channels.

[0072] As an implementation method, in step S4, the fusion mechanism can be replaced by a multi-task attention allocation model or a confidence regression network optimization weighted strategy.

[0073] As an implementation, in step S5, the recognition result supports non-exclusive fuzzy classification of unknown or borderline discharge patterns.

[0074] This method fundamentally differs from traditional approaches in that it eliminates the need for "black-box" modeling based on holistic features. Instead, it explicitly extracts structural semantic fragments (channels) from PD data. This strategy enhances the model's expressiveness, transferability, and tolerance to field defect signals through a strategy that integrates independent modeling with a hierarchical reasoning layer. Furthermore, this method naturally supports interpretable reasoning under "weak supervision," making it highly practical for engineering deployment.

[0075] Example 2

[0076] This embodiment further provides a partial discharge intelligent identification system based on collaborative reasoning, including:

[0077] Feature map construction module, used to build multi-scale feature information maps of partial discharge measurement data and extract high-order dependencies between dimensions;

[0078] The channel partitioning module is used to divide the high-dimensional feature space into multiple structurally consistent feature sub-channels through a structural decomposition algorithm. Each sub-channel corresponds to a local feature group related to the discharge mechanism.

[0079] A fuzzy inference module is used to independently construct a fuzzy inference model for each feature sub-channel and generate a channel-level discharge pattern sub-classifier;

[0080] The dynamic fusion module is used to dynamically adjust the fusion strategy based on the prediction confidence and perform structural consistency-driven fusion of multiple sub-classifier outputs.

[0081] As an implementation method, the dynamic fusion module has a built-in cross-channel prediction consistency verification mechanism, which triggers a confidence reweighting process for conflicting outputs.

[0082] The present invention uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be understood as limiting the present invention.

Claims

1. A collaborative reasoning-based intelligent identification method for partial discharge, characterized by: The following steps are involved: S1. Construct a multi-scale feature information map of partial discharge measurement data and extract high-order dependencies between dimensions; S2. Divide the high-dimensional feature space into multiple structurally consistent feature sub-channels using a structural decomposition algorithm, where each sub-channel corresponds to a local feature cluster related to the discharge mechanism; S3. Build a fuzzy inference model independently for each characteristic sub-channel to generate a channel-level discharge pattern sub-classifier; S4. Dynamically adjust the fusion strategy based on prediction confidence and perform structural consistency-driven fusion of multiple sub-classifier outputs; S5. Output the final partial discharge type identification result.

2. The method for intelligent identification of partial discharge based on collaborative reasoning according to claim 1, characterized in that: In step S1, the multi-scale feature information graph uses at least one of mutual information, GraphLasso algorithm, kernel correlation coefficient or Copula function to construct inter-dimensional dependencies.

3. The method for intelligent identification of partial discharge based on collaborative reasoning according to claim 1, characterized in that: In step S2, the structural decomposition algorithm is a sparse spectrum partitioning algorithm, a hierarchical Louvain clustering algorithm or an adaptive multi-channel clustering algorithm.

4. The method for intelligent identification of partial discharge based on collaborative reasoning according to claim 1, characterized in that: In step S3, the fuzzy reasoning model is constructed by automatically learning the fuzzy rule set and membership function structure.

5. The method for intelligent identification of partial discharge based on collaborative reasoning according to claim 1 is characterized in that: In step S3, the fuzzy inference model can be replaced by a sparse Bayesian regression model or an integrated fuzzy network model.

6. The method for intelligent identification of partial discharge based on collaborative reasoning according to claim 1, characterized in that: In step S4, the dynamic adjustment of the fusion strategy includes: Calculate the prediction confidence of each sub-classifier; Weighted fusion output according to confidence weight; Implement a fault-tolerant mechanism for inconsistencies between feature channels.

7. The method for intelligent identification of partial discharge based on collaborative reasoning according to claim 1, characterized in that: In step S4, the fusion mechanism can be replaced by a multi-task attention allocation model or a confidence regression network optimization weighted strategy.

8. The method for intelligent identification of partial discharge based on collaborative reasoning according to claim 1, characterized in that: In step S5, the recognition result supports non-exclusive fuzzy classification of unknown or borderline discharge patterns.

9. Partial discharge intelligent identification system based on collaborative reasoning, characterized by: include: Feature map construction module, used to construct multi-scale feature information maps of partial discharge measurement data and extract high-order dependencies between dimensions; The channel partitioning module is used to divide the high-dimensional feature space into multiple structurally consistent feature sub-channels through a structural decomposition algorithm. Each sub-channel corresponds to a local feature group related to the discharge mechanism. A fuzzy inference module is used to independently construct a fuzzy inference model for each feature sub-channel and generate a channel-level discharge pattern sub-classifier; The dynamic fusion module is used to dynamically adjust the fusion strategy based on the prediction confidence and perform structural consistency-driven fusion of multiple sub-classifier outputs.

10. The collaborative reasoning-based partial discharge intelligent identification system according to claim 9, characterized in that: The dynamic fusion module has a built-in cross-channel prediction consistency verification mechanism, which triggers a confidence re-weighting process for conflicting outputs.

Images