Helium leak detection method and system for switch cabinet

Through adaptive multi-decomposition algorithm and nonlinear manifold alignment technology, the nonlinear relationship and time-varying of multi-sensor data fusion and leakage pattern recognition in the prior art are solved by combining multi-sensor data fusion and time-varying. Characteristics problems achieve efficient and accurate helium leakage detection, and is suitable for large-scale switch cabinet systems.

CN118882947AActive Publication Date: 2024-11-01NANJING ZHENGRUI POWER TECH CO LTD

Patent Information

Application Number
CN202411382069.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-30
Publication Date
2024-11-01
Estimated Expiration
2044-09-30

AI Technical Summary

Technical Problem

The existing switch cabinet helium leakage detection technology has difficulty in processing nonlinear relationships in multi-sensor data fusion and leakage pattern recognition, time-varying characteristics are neglected, and computational efficiency and scalability are insufficient, making it difficult to achieve real-time monitoring and rapid response, especially in large-scale in a scale system.

Method used

Adaptive multi-resolution decomposition algorithm and nonlinear manifold alignment technology to process raw data, build multi-scale causal networks and time-varying information flow networks, combine dynamic graph embedding algorithms to generate multi-modal representations and spatiotemporal features, and perform graph topological anomalies Detection and parameter optimization.

Benefits of technology

It improves the high accuracy, high efficiency and adaptability of helium leak detection in large-scale switchgear systems, enhances the accuracy and reliability of detection, and can effectively deal with complex industrial environments and variable operating conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118882947A_ABST
    Figure CN118882947A_ABST
Patent Text Reader

Abstract

The invention discloses a helium leak detection method and system for a switch cabinet, and the method comprises the following steps: carrying out the self-adaptive multi-resolution decomposition, anomaly detection and nonlinear manifold alignment of original data, and extracting dynamic features; applying adaptive tensor decomposition and multi-scale causal network analysis to construct a fused multi-modal representation; carrying out nonlinear dynamic system reconstruction and continuous coherence analysis, and extracting spatio-temporal characteristics; constructing a multi-layer graph structure and generating a graph topology anomaly detection result by applying a dynamic graph embedding algorithm; and constructing a multi-objective optimization framework and a dynamic parameter adjustment strategy, and generating an optimized anomaly detection result. According to the method, through multi-scale analysis, nonlinear modeling and dynamic optimization, high-precision, high-efficiency and self-adaptive detection of helium leak detection of the switch cabinet is realized, and the accuracy and reliability of detection are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to detection technology, in particular to a helium leak detection method and system for a switch cabinet. Background Art

[0002] As a key equipment in the power system, the sealing performance of the switchgear directly affects the safety and reliability of the entire system. Helium leak detection technology plays an increasingly important role in switchgear sealing detection due to its high sensitivity and non-destructive detection characteristics. With the continuous expansion of the scale of the power grid and the improvement of the degree of intelligence, the switchgear faces a more complex operating environment and higher reliability requirements. Accurate and efficient helium leak detection methods can not only detect potential sealing problems in a timely manner and prevent serious accidents, but also optimize maintenance strategies, extend equipment life, and reduce operating costs. In addition, in the context of the transformation of the power system to clean energy, reducing the leakage of greenhouse gases such as SF6 has also become an important issue.

[0003] At present, the helium leak detection technology for switch cabinets has made certain progress. The traditional helium mass spectrometry leak detection method determines the leakage by measuring the concentration of helium, which has high sensitivity. However, this method often requires complex equipment and professional operation, and it is difficult to achieve large-scale field application. In recent years, the detection method based on pressure change has been widely used, and the leakage is inferred by monitoring the pressure change in the sealed chamber. This method is simple to operate, but the sensitivity is insufficient when detecting small leaks. At the same time, some researchers have tried to apply acoustic detection technology to helium leak detection, and locate the leak point by analyzing the sound wave characteristics generated by the leak. This method has a certain positioning capability, but it is easily interfered in a complex electromagnetic environment. In addition, some researchers have begun to try to introduce machine learning technology into helium leak detection analysis, and improve the accuracy and efficiency of detection by establishing data models.

[0004] However, the existing switch cabinet helium leak detection technology still faces some challenges. In terms of multi-sensor data fusion, existing methods often use simple linear combinations or weighted averages, which are difficult to effectively handle the nonlinear relationship and time-varying characteristics between different types of sensor data. In addition, in terms of leakage pattern recognition, most methods only focus on static features and ignore the dynamic evolution characteristics of the leakage process, resulting in insufficient recognition of complex leakage scenarios (such as multi-point micro-leakage or intermittent leakage). Existing anomaly detection algorithms are usually based on preset thresholds, lack adaptive capabilities, and are difficult to cope with dynamic changes in the switch cabinet operating environment.

[0005] In particular, when dealing with large-scale switchgear systems, the computational efficiency and scalability of existing methods are also challenged, making it difficult to achieve real-time monitoring and rapid response. These problems limit the application of helium leak detection technology in switchgear sealing detection. Summary of the invention

[0006] The purpose of the invention is to provide a helium leak detection method and system for a switch cabinet, in order to solve the above-mentioned problems existing in the prior art.

[0007] A technical solution, a helium leak detection method for a switch cabinet, comprises the following steps:

[0008] Step S1, acquiring raw data from the sensor in real time, organizing the raw data into an N-dimensional raw data tensor, applying an adaptive multi-resolution decomposition algorithm to the raw data tensor to obtain a decomposed data tensor; constructing a persistence graph based on the decomposed data tensor and calculating persistence homology, performing anomaly detection and repair, and generating a cleaned data tensor; performing nonlinear manifold alignment on the cleaned data tensor to obtain an aligned six-dimensional data tensor, extracting dynamic features from the aligned data tensor, and generating a dynamic feature matrix;

[0009] Step S2, obtaining the aligned data tensor and applying an adaptive tensor decomposition algorithm to it to obtain a low-dimensional representation; constructing a multi-scale causal network based on the low-dimensional representation to generate a multi-scale causal adjacency matrix; constructing a time-varying information flow network using the multi-scale causal adjacency matrix and the dynamic feature matrix; constructing a fused multimodal representation based on the low-dimensional representation, the multi-scale causal adjacency matrix and the time-varying information flow network;

[0010] Step S3, reading the fused multimodal representation, reconstructing the low-dimensional representation in the multimodal representation by a nonlinear dynamic system, and obtaining a phase space trajectory; calculating persistence coherence based on the phase space trajectory, and generating a persistence graph; extracting dynamic patterns from the phase space trajectory and the persistence graph, and obtaining a pattern matrix; combining the phase space trajectory, the persistence graph, and the pattern matrix into a spatiotemporal feature representation;

[0011] Step S4: read and construct a multi-layer graph structure based on the fused multimodal representation and spatiotemporal feature representation, apply a dynamic graph embedding algorithm to the multi-layer graph structure to obtain a low-dimensional representation of the node, calculate local and global topological features based on the low-dimensional representation of the node, and construct anomaly detection indicators; combine the multi-layer graph structure, the low-dimensional representation of the node, and the anomaly detection indicator into a graph topology anomaly detection result;

[0012] Step S5, read the graph topology anomaly detection results and construct a key parameter set and a multi-objective optimization framework to obtain a non-dominated solution set, build a dynamic parameter adjustment strategy based on the non-dominated solution set, and generate a time-varying optimal parameter set; combine the graph topology anomaly detection results and the time-varying optimal parameter set into an optimized anomaly detection result.

[0013] According to another aspect of the present application, a helium leak detection system for a switch cabinet is also provided, comprising:

[0014] at least one processor; and,

[0015] a memory communicatively connected to at least one of the processors; wherein,

[0016] The memory stores instructions executable by the processor, and the instructions are used to be executed by the processor to implement the helium leak detection method for a switch cabinet described in any one of the above technical solutions.

[0017] Beneficial effect: high-precision, high-efficiency and adaptive detection of helium leak detection in large-scale switch cabinet systems are improved, and the accuracy and reliability of detection are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flow chart of the present invention.

[0019] Figure 2 It is a flow chart of step S1 of the present invention.

[0020] Figure 3 It is a flow chart of step S2 of the present invention.

[0021] Figure 4 It is a flow chart of step S3 of the present invention.

[0022] Figure 5 It is a flow chart of step S4 of the present invention.

[0023] Figure 6 It is a flow chart of step S5 of the present invention. DETAILED DESCRIPTION

[0024] like Figure 1 As shown, the helium leak detection method for switch cabinet includes the following steps:

[0025] Step S1, acquiring raw data from the sensor in real time, organizing the raw data into an N-dimensional raw data tensor, applying an adaptive multi-resolution decomposition algorithm to the raw data tensor to obtain a decomposed data tensor; constructing a persistence graph based on the decomposed data tensor and calculating persistence homology, performing anomaly detection and repair, and generating a cleaned data tensor; performing nonlinear manifold alignment on the cleaned data tensor to obtain an aligned six-dimensional data tensor, extracting dynamic features from the aligned data tensor, and generating a dynamic feature matrix;

[0026] Step S2, obtaining the aligned data tensor and applying an adaptive tensor decomposition algorithm to it to obtain a low-dimensional representation; constructing a multi-scale causal network based on the low-dimensional representation to generate a multi-scale causal adjacency matrix; constructing a time-varying information flow network using the multi-scale causal adjacency matrix and the dynamic feature matrix; constructing a fused multimodal representation based on the low-dimensional representation, the multi-scale causal adjacency matrix and the time-varying information flow network;

[0027] Step S3, reading the fused multimodal representation, reconstructing the low-dimensional representation in the multimodal representation by a nonlinear dynamic system, and obtaining a phase space trajectory; calculating persistence coherence based on the phase space trajectory, and generating a persistence graph; extracting dynamic patterns from the phase space trajectory and the persistence graph, and obtaining a pattern matrix; combining the phase space trajectory, the persistence graph, and the pattern matrix into a spatiotemporal feature representation;

[0028] Step S4: read and construct a multi-layer graph structure based on the fused multimodal representation and spatiotemporal feature representation, apply a dynamic graph embedding algorithm to the multi-layer graph structure to obtain a low-dimensional representation of the node, calculate local and global topological features based on the low-dimensional representation of the node, and construct anomaly detection indicators; combine the multi-layer graph structure, the low-dimensional representation of the node, and the anomaly detection indicator into a graph topology anomaly detection result;

[0029] Step S5, read the graph topology anomaly detection results and construct a key parameter set and a multi-objective optimization framework to obtain a non-dominated solution set, build a dynamic parameter adjustment strategy based on the non-dominated solution set, and generate a time-varying optimal parameter set; combine the graph topology anomaly detection results and the time-varying optimal parameter set into an optimized anomaly detection result.

[0030] In this embodiment, a closed-loop intelligent detection framework is formed from data preprocessing, multimodal fusion, dynamic pattern extraction, anomaly detection to parameter optimization. The scheme can effectively handle high-dimensional data, nonlinear relationships and time-varying characteristics in the process of helium leak detection in switch cabinets, and improve the accuracy, sensitivity and reliability of detection. Especially in the face of complex industrial environments and changing operating conditions, the scheme shows excellent robustness and adaptability. Through multi-scale analysis and graph structure representation, the scheme can simultaneously deal with small leaks and large-scale leaks, and accurately locate the source of the leak. The dynamic optimization mechanism enables the system to continuously improve and adapt to equipment aging and environmental changes. In addition, the modular design of the scheme makes it have good scalability and versatility, which is not only suitable for switch cabinets, but also can be extended to the detection of other sealing equipment. Overall, the scheme provides strong technical support for preventive maintenance, fault diagnosis and reliability improvement of switch cabinets, and is expected to significantly reduce maintenance costs, improve equipment life and operational safety.

[0031] According to one aspect of the present application, step S1 specifically comprises:

[0032] Step S11, acquiring part of the raw data from at least two sensors in real time; calculating the rate of change of each sensor signal according to the part of the raw data, and determining the adaptive sampling frequency; resampling all sensor data using the adaptive sampling frequency; and organizing the resampled data into a raw data tensor;

[0033] Step S12, read the original data tensor, apply the adaptive multi-resolution decomposition algorithm to each sensor signal in the tensor, and use wavelet transform to perform multi-scale decomposition on the signal to obtain an initial decomposition set; apply empirical mode decomposition to the initial decomposition set to obtain a multi-scale component set; calculate the information entropy and computational complexity of each decomposition level; construct and optimize the objective function, select the optimal decomposition level; based on the optimal level, select the corresponding component from the multi-scale component set; combine the processed signals to generate a decomposition data tensor;

[0034] Step S13, reading the decomposed data tensor, constructing a time delay embedding vector for each decomposition level of each sensor signal in the decomposed data tensor; constructing a complex based on the embedding vector, calculating the persistence data, and obtaining a persistence graph; calculating anomaly metrics, and comparing the anomaly metrics with a preset threshold, marking abnormal data points; repairing the abnormal data points using a local linear regression method; reorganizing the repaired data to generate a cleaned data tensor;

[0035] Step S14, read the cleaned data tensor and use a dimensionality reduction algorithm to embed each decomposition level of each sensor signal into a high-dimensional manifold; calculate the geodesic distance between high-dimensional manifolds; set the sensor importance weight, construct the objective function, and use the optimization algorithm to minimize the objective function to obtain the optimal alignment parameter; based on the optimal alignment parameter, transform the data to obtain the aligned data tensor;

[0036] Step S15, read the aligned data tensor, apply the time-frequency analysis method to each decomposition level of each sensor signal in the data tensor; calculate the statistical moment of the time-frequency representation; extract geometric features; form a feature set; calculate the information entropy and mutual information between features of each feature; and construct a feature importance score, sort and select features based on the importance score; organize the selected features into a dynamic feature matrix.

[0037] In this embodiment, high-quality preprocessing and feature extraction of switch cabinet helium leak detection data are achieved. First, the original data is decomposed by an adaptive multi-resolution decomposition algorithm, which effectively separates signal components of different scales and improves the accuracy of subsequent analysis. The introduction of continuous coherence makes anomaly detection and repair more robust and can capture complex topological changes. The nonlinear manifold alignment technology solves the nonlinear deviation between multi-sensor data, so that data from different sources can be analyzed in a unified feature space. Dynamic feature extraction further enhances the method's ability to characterize time-varying patterns. The synergy of these technologies enables the method to extract high-quality, multi-scale, time-varying feature representations from the original data of switch cabinet helium leak detection, laying a solid foundation for subsequent anomaly detection and analysis. Especially in the complex electromagnetic environment of the switch cabinet, the method can effectively suppress noise interference, extract real helium leakage signal characteristics, and improve the sensitivity and reliability of detection.

[0038] According to one aspect of the present application, step S2 is specifically:

[0039] Step S21, read the aligned data tensor and dynamic feature matrix; initialize the decomposition rank parameter; apply Tucker decomposition to the data tensor to obtain a core tensor and a factor matrix; calculate the Tucker reconstruction error; apply CP decomposition to the data tensor to obtain a factor matrix; calculate the CP reconstruction error, compare the errors of the two decomposition methods, and select the method with the smaller error; reconstruct the data based on the selected decomposition method to obtain a low-dimensional representation; reshape the low-dimensional representation into a low-dimensional data tensor;

[0040] Step S22, read the low-dimensional data tensor; construct a time scale set; for each pair of sensor signals in the low-dimensional data tensor, calculate the multi-scale Granger causality at each time scale; perform wavelet transform on the low-dimensional data tensor at each scale to obtain wavelet coefficients; construct a vector autoregression model; calculate the F statistic; calculate the multi-scale causal strength; organize all the multi-scale causal strengths into a multi-scale causal adjacency matrix;

[0041] Step S23, read the multi-scale causal adjacency matrix and the dynamic feature matrix; for each pair of sensor signals in the dynamic feature matrix, calculate the conditional entropy at each time point; use the kernel density estimation method to estimate the probability from the dynamic feature matrix; calculate the information flow; construct a time-varying information flow network, using sensors as nodes and information flow as edge weights; apply the community detection algorithm to the time-varying information flow network to identify the dynamic community structure; calculate the centrality index of each node in the time-varying information flow network; represent the time-varying information flow network as a combination of the adjacency matrix and the node attribute matrix; combine the low-dimensional data tensor, the multi-scale causal adjacency matrix and the time-varying information flow network into a fused multimodal representation.

[0042] In this embodiment, a multimodal and multiscale representation of switch cabinet helium leak detection data is realized. The adaptive tensor decomposition algorithm can effectively reduce the data dimension while retaining key information. The construction of a multiscale causal network reveals the causal relationship between helium leakage signals at different scales, which helps to understand the leakage mechanism. The introduction of a time-varying information flow network captures the dynamic changes in the leakage process. The combination of these technologies enables the method to comprehensively characterize the complex dynamic behavior of the switch cabinet helium leak detection process. Especially when dealing with large-scale, high-dimensional detection data, the method exhibits excellent computational efficiency and scalability. Through the fusion of multimodal representations, the method can simultaneously utilize data from multiple sensors (such as pressure, temperature, electromagnetic field, etc.) to comprehensively evaluate the sealing performance of the switch cabinet, thereby improving the accuracy and reliability of the detection. In addition, multiscale analysis enables the method to capture both small leaks and large-scale leaks at the same time, meeting different degrees of sealing requirements.

[0043] According to one aspect of the present application, step S3 is specifically:

[0044] Step S31, read the low-dimensional data tensor in the fused multimodal representation; calculate the mutual information function for each sensor component in the low-dimensional data tensor; select the time delay that makes the mutual information function reach the local minimum for the first time as the optimal time delay; use the error prediction method to estimate the optimal embedding dimension; construct a candidate embedding vector; for each candidate embedding dimension, use the k-nearest neighbor method to predict the future value and calculate the prediction error; when the prediction error no longer decreases significantly, use the dimension as the optimal embedding dimension; construct a phase space trajectory using the optimal time delay and the optimal embedding dimension; repeat this process for all sensor components in the low-dimensional data tensor to obtain a complete phase space trajectory;

[0045] Step S32, read the complete phase space trajectory; apply a sliding window to the phase space trajectory to obtain a time series set; construct a Vietoris-Rips complex for each time series in the time series set; calculate the multidimensional persistence homology of the complex to obtain a persistence graph; combine the persistence graphs of all windows into a time-varying persistence graph; calculate the statistical features of the time-varying persistence graph, including persistence entropy and Betti number curves; define a topological complexity measure; combine the time-varying persistence graph and the topological complexity measure into a topological feature representation;

[0046] Step S33, read the complete phase space trajectory and topological feature representation; apply continuous wavelet transform to the phase space trajectory; calculate the wavelet energy spectrum; identify significant peaks in the wavelet energy spectrum to obtain a candidate pattern set; for each candidate pattern in the candidate pattern set, calculate its duration and energy; define a pattern importance index; sort the candidate patterns according to the pattern importance index and select the main dynamic pattern; for each selected main dynamic pattern, extract its time-frequency characteristics and shape characteristics; combine the extracted features with the topological features of the corresponding time period to form a pattern descriptor; organize all pattern descriptors into a pattern matrix; combine the phase space trajectory, the time-varying persistence diagram and the pattern matrix into a spatiotemporal feature representation.

[0047] In this embodiment, the complex dynamic patterns in the switch cabinet helium leak detection data are deeply mined through nonlinear dynamic reconstruction and persistence coherence analysis. Phase space reconstruction technology can recover the intrinsic dynamics of the system from time series data and effectively capture the nonlinear and non-stationary characteristics of the helium leakage process. The calculation of persistence coherence provides in-depth insights into the topological structure of the data and can identify stable leakage patterns and transient anomalies. Dynamic pattern extraction further enhances the method's ability to characterize time-varying behaviors. The combined application of these technologies enables the method to extract rich spatiotemporal feature representations from the switch cabinet helium leak detection data. Especially when dealing with complex leakage situations, such as multi-point micro-leakage or intermittent leakage, the method exhibits excellent recognition capabilities. By analyzing the phase space trajectory and persistence diagram, the leakage source can be accurately located, the leakage rate can be estimated, and even the development trend of the leakage can be predicted. This provides strong support for preventive maintenance and fault diagnosis of switch cabinets.

[0048] According to one aspect of the present application, step S4 is specifically:

[0049] Step S41, read the fused multimodal representation and spatiotemporal feature representation; define a layer set, each layer represents a different interaction type or scale; for each layer in the layer set, construct a node set, including sensor nodes and virtual nodes generated by dynamic patterns in the spatiotemporal feature representation; construct an edge set based on similarities in low-dimensional data tensors, causal relationships in multi-scale causal adjacency matrices, and information flows in time-varying information flow networks; define a node attribute set, including features in low-dimensional data tensors, dynamic features in phase space trajectories, and topological features in time-varying persistence graphs; calculate inter-layer coupling strength; construct an inter-layer edge set to connect nodes in different layers that represent the same entity or are highly correlated; integrate information from all layers to form a multi-layer graph structure;

[0050] Step S42, read the multi-layer graph structure; initialize the node embedding matrix; define the intra-layer loss function based on the adjacency matrix of each layer; define the inter-layer loss function based on the inter-layer coupling strength; define the attribute preservation loss function based on the node attribute set; construct the overall objective function, combining the intra-layer loss, inter-layer loss and attribute preservation loss; use the stochastic gradient descent method to optimize the overall objective function; update the node embedding matrix until the objective function converges or reaches the maximum number of iterations; output the final node embedding;

[0051] Step S43, read node embedding and multi-layer graph structure; for each node in the multi-layer graph structure, calculate local topological features; extract the k-hop neighborhood of the node; calculate the local clustering coefficient; calculate the local heterogeneity coefficient; calculate the local page rank; combine these features to obtain local topological features; calculate global topological features; apply the spectral clustering algorithm to the node embedding to obtain the community affiliation of the node; calculate the centrality index of the node, including eigenvector centrality and betweenness centrality; calculate the structural role of the node; combine these features to obtain global topological features; define anomaly scores, combining local and global topological features; combine the multi-layer graph structure, node embedding and anomaly scores into a graph topology anomaly detection result.

[0052] By constructing a multi-layer graph structure and applying a dynamic graph embedding algorithm, efficient anomaly detection of switchgear helium leak detection data is achieved. The multi-layer graph structure can simultaneously represent relationships of different types and scales, such as spatial relationships between sensors, similarities of measured values, etc. The dynamic graph embedding algorithm maps complex graph structures into low-dimensional space, improving computational efficiency. The calculation of local and global topological features makes anomaly detection more comprehensive and robust. The combination of these technologies enables the method to quickly and accurately identify abnormal patterns in the switchgear helium leak detection process. In particular, the method shows excellent scalability when dealing with large-scale switchgear systems. By analyzing the low-dimensional representation and topological features of the nodes, abnormal sensors or abnormal areas can be quickly located, shortening the fault diagnosis time. In addition, the use of the graph structure also enables the method to consider the complex structural relationships inside the switchgear, such as the connectivity between different compartments, thereby providing more accurate leak location results.

[0053] According to one aspect of the present application, step S5 is specifically:

[0054] Step S51, read the graph topology anomaly detection result and the key parameter set used in all previous steps; define the detection accuracy objective function, based on the area under the curve of the receiver operating characteristic curve of the anomaly score; define the computational efficiency objective function, based on the total computing time and the maximum allowed computing time; construct a set of constraints, including parameter value ranges and resource restrictions; initialize the population, each individual represents a set of parameter configurations; for each individual in the population, calculate the objective function value; use non-dominated sorting to stratify the population and obtain a non-dominated level; calculate the crowding distance; perform selection, crossover and mutation operations based on the non-dominated level and crowding distance to generate a new population; repeat the steps from initializing the population to generating a new population until the maximum number of iterations is reached or convergence occurs; output the final non-dominated solution set;

[0055] Step S52, read the non-dominated solution set and the graph topology anomaly detection results; define performance indicators, combining precision, recall and false positive rate; initialize the Q function, the state represents the current parameter configuration, and the action represents the parameter adjustment strategy; define the state transition function, which represents the probability of transferring from one state to another after executing an action; define the reward function based on the change in the performance indicator; use the Q-learning algorithm to update the Q function; at each time step, select an action with the ε-greedy strategy; execute the selected action, update the parameter configuration, and obtain a new state; observe the reward and update the Q function; based on the updated Q function, generate a time-varying optimal parameter set; combine the graph topology anomaly detection result and the time-varying optimal parameter set into an optimized anomaly detection result.

[0056] Through multi-objective optimization and dynamic parameter adjustment, the adaptive optimization of the switch cabinet helium leak detection method is realized. The multi-objective optimization framework can simultaneously consider multiple objectives such as detection accuracy and computational efficiency to find the best balance point. The generation of non-dominated solution sets provides a series of Pareto optimal solutions, which provides flexibility for decision-making. The dynamic parameter adjustment strategy enables the method to adapt to changes in the operating status of the switch cabinet and always maintain the best performance. The synergy of these technologies enables the method to always maintain efficient and accurate detection capabilities under different working conditions. Especially in long-term monitoring tasks, the method shows excellent stability and adaptability. By continuously optimizing parameter settings, the method can adapt to the influence of factors such as switch cabinet aging and environmental changes, and maintain the consistency of detection performance. In addition, the generation of time-varying optimal parameter sets also provides an important reference for the intelligent operation and maintenance of switch cabinets, which can guide the dynamic adjustment of key parameters such as detection frequency and sensitivity, maximizing the detection effect while minimizing resource consumption.

[0057] According to one aspect of the present application, in step S12, specifically:

[0058] Step S121, read the original data tensor; apply the sliding window technology to each sensor signal in the data tensor; use the improved box counting method to estimate the local fractal dimension in each window; calculate the fractal spectrum of the signal to obtain the change of the fractal dimension over time; store the fractal dimension time series as a fractal feature matrix;

[0059] Step S122, read the fractal feature matrix; use the dynamic programming algorithm to find the optimal segmentation point on the fractal dimension time series; each segment corresponds to a characteristic scale; calculate the average fractal dimension of each segment; according to the average fractal dimension, adaptively select the decomposition basis function family; store the selected basis function family and the corresponding scale information as a scale selection matrix;

[0060] Step S123, read the original data tensor and the scale selection matrix; decompose each sensor signal in the data tensor at each adaptively selected scale; use the basis function family corresponding to the scale to decompose the signal; calculate the decomposition coefficient at each scale; combine the decomposition coefficients of all scales into a multi-scale decomposition tensor;

[0061] Step S124, read the multi-scale decomposition tensor; apply kernel principal component analysis to extract nonlinear features for the coefficients of each scale in the decomposition tensor; use radial basis function kernel for nonlinear mapping; select the principal component that can explain 90% of the variance; store the extracted nonlinear features as a feature tensor;

[0062] Step S125, read the feature tensor and the original data tensor; use the particle swarm optimization algorithm to find the optimal reconstruction weight; the objective function is the weighted sum of the reconstruction error and the information entropy; for each time point, based on the current optimal weight, reconstruct the features in the feature tensor into signals; combine the reconstructed signals into a reconstructed data tensor;

[0063] Step S126, reading the reconstructed data tensor and the original data tensor; calculating the residual between the reconstructed data tensor and the original data tensor; processing the residual using an adaptive Kalman filter; dynamically adjusting the parameters of the filter according to the statistical characteristics of the residual; adding the filtered residual to the reconstructed data tensor to obtain a refined reconstructed data tensor;

[0064] Step S127, read the finely reconstructed data tensor and the original data tensor; calculate multiple evaluation indicators, including root mean square error, structural similarity index and information retention rate; use fuzzy comprehensive evaluation method to integrate multiple indicators into a single quality score; if the quality score is lower than the preset threshold, return to step S122 for parameter adjustment and re-decomposition;

[0065] Step S128, read the finely reconstructed data tensor, the multi-scale decomposition tensor and the feature tensor; align the size and structure of the finely reconstructed data tensor with the original data tensor; combine the multi-scale decomposition tensor and the feature tensor to generate an enhanced feature tensor; merge the finely reconstructed data tensor and the enhanced feature tensor to form the final decomposition data tensor.

[0066] Through signal processing technology, high-quality decomposition and enhancement of the raw data of switchgear helium leak detection are achieved. The adaptive multi-resolution decomposition algorithm can automatically select the optimal decomposition scale according to the local characteristics of the signal and effectively separate different frequency components. The introduction of fractal analysis and kernel principal component analysis further improves the nonlinear ability of feature extraction. The application of adaptive Kalman filter effectively suppresses noise interference. The synergy of these technologies enables this method to extract high-quality, multi-scale feature representations from complex switchgear helium leak detection signals. In particular, this method shows excellent performance when processing non-stationary, multi-scale leakage signals. Through adaptive decomposition, both fast-changing transient leakage and slowly evolving long-term leakage trends can be captured simultaneously. The introduction of fractal analysis enables the method to identify self-similar structures in the signal, which is particularly effective for detecting tiny periodic leaks. Kernel principal component analysis enhances the method's ability to extract nonlinear features and can capture complex leakage patterns. In addition, the quality assessment and parameter adaptive adjustment mechanism ensure the reliability and consistency of the decomposition results, providing a high-quality data foundation for subsequent analysis.

[0067] According to one aspect of the present application, step S14 is specifically:

[0068] Step S141, read the cleaned data tensor; apply the t-SNE algorithm to perform initial dimensionality reduction on each decomposition level of each sensor signal in the data tensor; use cosine similarity as a distance metric; generate an initial low-dimensional embedding matrix;

[0069] Step S142, read the initial low-dimensional embedding matrix; use the dynamic time warping algorithm to calculate the similarity between time series; construct an adaptive Gaussian kernel function based on the DTW distance; calculate the kernel matrix; store the kernel matrix as a similarity matrix;

[0070] Step S143, read the similarity matrix and the initial low-dimensional embedding matrix; construct the graph Laplacian matrix; define the manifold consistency objective function, combining local preservation and global alignment; optimize the objective function using the alternating direction multiplier method; update the low-dimensional embedding matrix; store the optimized low-dimensional embedding matrix as the aligned embedding matrix;

[0071] Step S144, read the alignment embedding matrix; use the recurrent neural network RNN ​​to learn the time-varying mapping function; the input of the RNN is the embedding vector of the adjacent time step, and the output is the predicted embedding of the next time step; minimize the prediction error and manifold consistency loss; store the learned RNN model parameters as the time-varying mapping model;

[0072] Step S145, read the time-varying mapping model and the alignment embedding matrix; use the learned RNN model to interpolate or extrapolate the data at missing or future time points; generate a complete time series embedding; merge the interpolation and extrapolation results with the original alignment embedding matrix to form an extended embedding matrix;

[0073] Step S146, read the extended embedding matrix; use wavelet transform to perform multi-scale decomposition on the embedding; apply the local linear embedding algorithm at each scale; merge the embedding results of different scales by weighted summation; dynamically adjust the weights using an adaptive fuzzy inference system; generate a fused multi-scale embedding matrix;

[0074] Step S147, read the fused multi-scale embedding matrix; construct continuous homology features; use continuous homology information to guide manifold deformation; minimize Wasserstein distance to maintain topological structure; apply discrete external differential operator for smoothing; generate topology-preserving embedding matrix;

[0075] Step S148, reading the topology-preserving embedding matrix and the original data tensor; calculating the geodesic distance preservation rate before and after the manifold alignment; evaluating the degree of preservation of the local neighborhood structure; using the mutual information criterion to quantify the degree of alignment between different sensor signals; and generating an alignment quality score by combining multiple indicators;

[0076] Step S149, read the topology-preserving embedding matrix, alignment quality score and original data tensor; based on the alignment quality score, adaptively adjust the embedding weight; map the weighted embedding back to the original data space; fuse it with the original data tensor to generate the final aligned data tensor.

[0077] In this embodiment, nonlinear alignment of multi-sensor data of the switch cabinet is achieved. The combination of t-SNE algorithm and dynamic time warping effectively solves the dual challenges of high-dimensional data dimensionality reduction and time series alignment. The introduction of the graph Laplacian matrix takes into account the local structure of the data and enhances the accuracy of the alignment. The application of recurrent neural network realizes the learning of time-varying mapping functions, so that the method can adapt to dynamically changing alignment relationships. The comprehensive application of these technologies enables the method to eliminate nonlinear deviations between multiple sensors while maintaining the essential characteristics of the data. In particular, when dealing with the complex electromagnetic environment and temperature distribution inside the switch cabinet, the method exhibits excellent performance. Through precise data alignment, the systematic errors caused by factors such as position and sensitivity between different sensors can be eliminated, which improves the accuracy of subsequent analysis. Multi-scale alignment and topology retention mechanisms ensure that the alignment process does not lose key structural information, which is crucial for identifying complex leakage patterns. In addition, the adaptive weight adjustment mechanism enables the method to dynamically balance the contributions of different sensors and maximize information utilization efficiency.

[0078] According to one aspect of the present application, step S21 is specifically:

[0079] Step S211, read the aligned data tensor and dynamic feature matrix; standardize the data tensor; calculate the modal correlation matrix of the tensor; group the modal using the spectral clustering algorithm; generate the modal grouping information matrix;

[0080] Step S212, read the modality grouping information matrix; for each modality group, call a predetermined kernel function; use Gaussian process regression to automatically select kernel function hyperparameters; generate a kernel function parameter matrix; apply the kernel function to the original data tensor to generate a kernelized data tensor;

[0081] Step S213, read the kernelized data tensor; use the sliding window technique to apply tensor singular value decomposition to each time window; calculate the singular value decay curve; use the Bayesian information criterion to estimate the optimal rank of each window; use the long short-term memory network to predict the rank of future time windows; generate a dynamic rank estimation sequence;

[0082] Step S214, read the kernelized data tensor and the dynamic rank estimation sequence; apply different tensor decomposition methods to each modality group: use Tucker decomposition for dense modality groups; use CP decomposition for sparse modality groups; use tensor column selection algorithm for mixed modality groups; perform decomposition operations in combination with dynamic rank information; generate decomposition factor matrix sets and core tensors;

[0083] Step S215, read the decomposition factor matrix set; apply structured sparse regularization to each factor matrix; use the proximal gradient descent algorithm to optimize the regularization objective function; introduce a temporal smoothing constraint to ensure the temporal continuity of the factor; generate a regularized factor matrix set;

[0084] Step S216, read the regularized factor matrix set and core tensor; call the adaptive fusion network based on the attention mechanism; use reinforcement learning to train the fusion strategy, and the reward function combines the reconstruction error and the information retention rate; generate the optimal fusion weight matrix;

[0085] Step S217, read the fusion weight matrix and the new data transmitted in real time; adopt the incremental update rule based on the random approximation algorithm; update the factor matrix and the core tensor online; use the exponential sliding average to maintain the long-term statistical information; generate the updated decomposition result;

[0086] Step S218, read the updated decomposition result and the original data tensor; calculate the reconstruction error tensor; perform principal component analysis on the error tensor to extract the main error mode; call the pre-built error predictor based on wavelet neural network; generate an error compensation model;

[0087] Step S219, read the updated decomposition result, the error compensation model and the original data tensor; reconstruct the data tensor based on the decomposition result to obtain a low-dimensional representation.

[0088] In this embodiment, efficient dimensionality reduction and feature extraction of switch cabinet helium leak detection data are achieved. The introduction of kernel function enhances the method's ability to model nonlinear relationships. Dynamic rank estimation technology can adaptively select the optimal decomposition rank, avoiding information loss and overfitting. Structured sparse regularization promotes the interpretability and stability of factors. The synergy of these technologies enables the method to extract low-dimensional but information-rich representations from high-dimensional switch cabinet detection data. In particular, the method exhibits excellent performance when processing the fusion of data from multiple sensors (such as pressure, temperature, acoustics, etc.). Through kernelization processing, complex nonlinear relationships between different types of data can be effectively captured, such as the effect of temperature on the sensitivity of pressure sensors. Dynamic rank estimation enables the method to adapt to changes in the operating state of the switch cabinet, such as the transition from normal operation to a faulty state. Sparse regularization helps to identify the most critical feature combinations, improving the interpretability and noise resistance of the method. In addition, the incremental update and error compensation mechanism enable the method to process real-time data streams, providing the possibility for online monitoring of switch cabinets.

[0089] According to another aspect of the present application, a helium leak detection system for a switch cabinet is also provided, comprising:

[0090] at least one processor; and,

[0091] a memory communicatively connected to at least one of the processors; wherein,

[0092] The memory stores instructions executable by the processor, and the instructions are used to be executed by the processor to implement the helium leak detection method for a switch cabinet described in any one of the above technical solutions.

[0093] According to one aspect of the present application, step S23 is specifically:

[0094] Step S231, read the multi-scale causal adjacency matrix and the dynamic feature matrix. Apply empirical mode decomposition (EMD) to each time series in the dynamic feature matrix. Generate multiple intrinsic mode functions (IMFs). Calculate the Hilbert transform for each IMF to obtain the instantaneous frequency and amplitude. Store the decomposition results as a multi-scale time series tensor.

[0095] Step S232, read the multi-scale time series tensor. For each scale, calculate the traditional Shannon entropy. Introduce Renyi entropy and Tsallis entropy as supplementary information metrics. Use the particle swarm optimization algorithm to adaptively select the optimal α parameter. Construct a comprehensive information metric index. Store the result as an adaptive information metric matrix.

[0096] Step S233, read the adaptive information metric matrix and the multi-scale causal adjacency matrix. Use the time-varying Granger causality test to infer the causal relationship in each time window. Introduce transfer entropy as a supplementary measure of nonlinear causality. Use the adaptive LASSO method for sparse processing. Generate a dynamic causal network sequence.

[0097] Step S234: read the dynamic causal network sequence. Apply the maximum flow algorithm to identify the main information flow paths. Introduce the concept of marginal information gain to quantify the information contribution of each path. Use the PageRank algorithm to evaluate the importance of nodes in information dissemination. Generate an information flow path importance matrix.

[0098] Step S235, read the dynamic causal network sequence. Calculate the time-varying network degree distribution, clustering coefficient and betweenness centrality. Introduce the spectral clustering method to identify the dynamic community structure. Use the Infomap algorithm to analyze the diffusion pattern of information flow in the network. Generate a network topology feature vector sequence.

[0099] Step S236: read the information flow path importance matrix and the network topology feature vector sequence. Design a hierarchical attention mechanism to adaptively integrate information flows of different scales. Use a graph neural network to capture high-order dependencies between nodes. Apply a dynamic time warping (DTW) algorithm to align information flows of different scales. Generate a multi-level information flow representation.

[0100] Step S237, read the multi-level information flow representation. Use the adaptive threshold method to construct a time-varying adjacency matrix. Introduce the concept of Forman-Ricci curvature to quantify the information flow resistance in the network. Apply the spectral method to optimize the network structure. Generate an optimized time-varying information flow network.

[0101] Step S238, read the optimized time-varying information flow network. Apply the multi-scale skeleton network extraction algorithm to identify the core structure of the network. Use the Louvain method to perform dynamic community detection. Introduce information theory-based metrics to evaluate community stability. Analyze community splitting, merging, and evolution patterns. Generate a dynamic community structure sequence.

[0102] Step S239, read the dynamic community structure sequence and the time-varying information flow network. Design a time series prediction model based on the graph attention network (GAT). Introduce an adaptive sampling strategy to balance short-term and long-term dependencies. Use adversarial training to improve the generalization ability of the model. Generate an information flow prediction model.

[0103] Step S2310, read the information flow prediction model, the time-varying information flow network and the dynamic community structure sequence. Calculate multiple evaluation indicators, including prediction accuracy, network stability and information flow efficiency. Generate an overall score using a fuzzy comprehensive evaluation method. Combine the time-varying information flow network, the dynamic community structure and the prediction model into the final dynamic information flow analysis result. Output this comprehensive result as the output of the multi-scale adaptive dynamic information flow analysis.

[0104] In this embodiment, a high-precision characterization of complex dynamic information flows in switchgear helium leak detection data is achieved. The combination of empirical mode decomposition and adaptive information metric provides in-depth analysis capabilities for multi-scale non-stationary signals. The introduction of time-varying Granger causality and transfer entropy enhances the method's ability to identify nonlinear causal relationships. The construction of dynamic community detection and information flow prediction models provides insights into the long-term evolution of system behavior. The combined application of these technologies enables the method to comprehensively and accurately describe the complex information propagation and interaction processes inside the switchgear. In particular, the method exhibits excellent performance when analyzing large switchgear systems or complex multi-point leakage scenarios. Through multi-scale information flow analysis, rapid local leakage events and slowly evolving global sealing performance degradation can be captured simultaneously. The construction of a dynamic causal network helps to understand the mutual influence between different components, such as how leakage in one area affects the pressure distribution in other areas. Community structure analysis can identify key subsystems or weak links in the system and provide guidance for directional maintenance. In addition, the introduction of the information flow prediction model enables the method to have the ability of predictive maintenance and can identify potential leakage risks in advance. The application of adaptive sampling and adversarial training further improves the generalization and robustness of the model, enabling it to cope with the dynamic changes in the switchgear operating environment. Overall, this step not only provides a deep understanding of the switchgear helium leak detection process, but also provides strong data support for fault diagnosis, risk assessment and maintenance decision-making, which is expected to significantly improve the operating reliability and maintenance efficiency of the switchgear.

[0105] According to one aspect of the present application, step S31 is specifically:

[0106] Step S311: read the low-dimensional data tensor in the fused multimodal representation, apply the wavelet threshold denoising method, use the adaptive threshold selection algorithm, calculate the signal-to-noise ratio and denoising effect, and generate the denoised data tensor.

[0107] Step S312: read the denoised data tensor. Calculate sample entropy, fuzzy entropy and permutation entropy. Use sliding window technology to obtain the change of entropy over time. Apply principal component analysis to reduce the dimension of entropy features. Generate a multi-scale entropy feature matrix.

[0108] Step S313, read the denoised data tensor and multi-scale entropy feature matrix. Calculate the self-mutual information function. Use the adaptive kernel density estimation method. Find the first local minimum of the self-mutual information function. Introduce the entropy-based correction factor. Generate the initial time delay estimate.

[0109] Step S314: read the initial time delay estimate and the denoised data tensor. Apply the pseudo nearest neighbor algorithm. Use the recursive graph analysis method. Calculate the relevant dimensions. Combine the results of these methods to generate a candidate set of embedding dimensions.

[0110] Step S315: read the embedding dimension candidate set, the initial time delay estimate, and the denoised data tensor. Reconstruct the phase space for each candidate embedding dimension. Calculate the maximum Lyapunov exponent. Use the Wolf algorithm and the Rosenstein algorithm and take the weighted average of the two. Generate a Lyapunov exponent matrix.

[0111] Step S316: read the Lyapunov exponent matrix, the embedding dimension candidate set, and the initial time delay estimate. Calculate the deterministic test statistic. Evaluate the fractal dimension of the reconstructed attractor. Use the mutual information rate criterion. Combine these indicators to construct a reconstruction quality score. Generate a reconstruction quality score matrix.

[0112] Step S317, read the reconstruction quality score matrix, Lyapunov index matrix and multi-scale entropy feature matrix. Construct a multi-objective optimization problem, the objectives include maximizing the reconstruction quality, maximizing the Lyapunov index, and minimizing the computational complexity. Use the improved NSGA-III algorithm to solve and generate a Pareto optimal solution set.

[0113] Step S318: Read the Pareto optimal solution set. Use a fuzzy logic-based decision system. Consider the dynamic characteristics of the system and computing resource constraints. Adaptively select the optimal time delay and embedding dimension combination. Generate the optimal reconstruction parameter set.

[0114] Step S319: read the optimal reconstruction parameter set and the denoised data tensor. Reconstruct the phase space using the selected time delay and embedding dimension. Apply kernel principal component analysis for manifold learning. Introduce diffusion map method to capture nonlinear structure. Generate reconstructed phase space trajectory.

[0115] Step S3110, read the reconstructed phase space trajectory. Calculate the recursive graph and cross-recursive graph of the trajectory. Analyze the periodicity and intermittency of the system. Use high-order spectrum analysis method to detect weak chaos. Construct a prediction model based on echo state network (Echo State Network). Generate system stability report and short-term prediction results.

[0116] Step S3111, read the system stability report, short-term prediction results and reconstructed phase space trajectory. Calculate the reconstruction fidelity index. Evaluate the performance of the prediction model. Generate a comprehensive evaluation report. Combine the reconstructed phase space trajectory, the optimal reconstruction parameter set, the prediction model and the evaluation report into the final nonlinear dynamic reconstruction result. Output this comprehensive result as the output of the adaptive multi-standard nonlinear dynamic reconstruction.

[0117] In this embodiment, the complex dynamic behavior in the switch cabinet helium leak detection data is accurately characterized. The combination of adaptive denoising and multi-scale entropy analysis provides a high-quality, information-rich data foundation for subsequent analysis. The adaptive selection of optimal time delay and embedding dimension ensures the accuracy of phase space reconstruction. The calculation and deterministic test of Lyapunov exponents provide in-depth insights into the dynamic characteristics of the system. The synergy of these technologies enables the method to recover the intrinsic dynamic structure of the system from the time series data of switch cabinet helium leak detection. In particular, the method exhibits excellent performance when dealing with complex nonlinear and non-stationary leakage processes. Through phase space reconstruction, seemingly random time series can be transformed into structured trajectories, thereby revealing potential leakage patterns. The calculation of the maximum Lyapunov exponent helps to evaluate the stability and predictability of the system, which is crucial for distinguishing normal fluctuations from abnormal leakage. The introduction of multi-objective optimization ensures the global optimality of the reconstruction parameters and balances the reconstruction quality, computational efficiency and model complexity. In addition, the application of kernel principal component analysis and diffusion map methods further enhances the method's ability to extract nonlinear features and helps to identify subtle leak precursors. The construction of the prediction model not only provides short-term prediction capabilities, but also provides an important basis for the health status assessment of the switchgear.

[0118] According to one aspect of the present application, step S33 is specifically:

[0119] Step S331, read the complete phase space trajectory and topological feature representation. Apply continuous wavelet transform and use adaptive wavelet basis function. Calculate the time-frequency energy spectrum. Introduce Wigner-Ville distribution to enhance the time-frequency resolution. Generate a multi-resolution time-frequency representation matrix.

[0120] Step S332: read the multi-resolution time-frequency representation matrix, apply the adaptive threshold segmentation algorithm, use morphological operations to optimize the segmentation results, calculate the shape descriptor of the segmented area, and generate an initial pattern candidate set.

[0121] Step S333, read the initial pattern candidate set and the multi-resolution time-frequency representation matrix. Extract time domain features (mean, variance, skewness, kurtosis). Calculate frequency domain features (power spectral density, spectral entropy). Extract time-frequency joint features (instantaneous frequency, group delay). Use an adaptive feature selection algorithm based on the maximum relevance minimum redundancy (mRMR) criterion. Generate a dynamic feature vector set.

[0122] Step S334, read the dynamic feature vector set and topological feature representation. Calculate the continuous homology features (Betti number curve, continuous graph). Extract the substitution topological features (Euler characteristic curve). Design a feature fusion network based on the attention mechanism. Learn the optimal feature combination strategy. Generate a fusion feature matrix.

[0123] Step S335: read the fusion feature matrix. Apply the adaptive spectral clustering algorithm to automatically determine the optimal number of clusters. Introduce a multi-view learning framework to integrate clustering results of different scales. Use the consistency matrix to evaluate clustering stability. Generate multi-scale pattern clustering results.

[0124] Step S336, read the multi-scale pattern clustering results. Apply the sequential pattern mining algorithm (such as PrefixSpan). Introduce time constraints and consider the periodicity and intermittency of the pattern. Use the Markov decision process to model the pattern transition. Calculate the pattern transition probability matrix. Generate a time series pattern library.

[0125] Step S337, read the time series pattern library and fusion feature matrix. Design an anomaly detector based on isolation forest. Introduce dynamic time warping (DTW) to measure pattern similarity. Calculate the anomaly score of the pattern. Use extreme value theory to determine the adaptive threshold. Generate an abnormal pattern report.

[0126] Step S338, read the time series pattern library and abnormal pattern report. Construct the pattern evolution diagram. Analyze the splitting, merging and transformation process of the pattern. Use complex network theory to quantify the importance of the pattern. Apply information flow theory to analyze the causal relationship between patterns. Generate a pattern evolution analysis report.

[0127] Step S339, read the pattern evolution analysis report and fusion feature matrix. Use a deep time series model (such as TCN or Transformer). Introduce an attention mechanism to capture long-term dependencies. Use Bayesian optimization to automatically adjust model hyperparameters. Apply ensemble learning to improve prediction stability. Generate a pattern prediction model.

[0128] Step S3310: Read the pattern prediction model and pattern evolution analysis report. Design a semantic mapping system based on the knowledge graph. Convert numerical features into interpretable semantic descriptions. Apply causal reasoning to explain the relationship between patterns. Generate a pattern semantic interpretation report.

[0129] Step S3311, read the pattern semantic interpretation report, pattern prediction model and abnormal pattern report. Calculate the completeness, consistency and novelty indicators of pattern extraction. Evaluate the performance of the prediction model (accuracy, recall, F1 score). Design an interactive visualization interface to display the dynamic evolution process of the pattern. Generate a comprehensive evaluation report and visualization results.

[0130] Step S3312: Read the comprehensive evaluation report, visualization results, pattern prediction model, and pattern semantic interpretation report. Integrate all analysis results to generate a final dynamic pattern extraction report. Combine the phase space trajectory, time-varying persistence graph, and pattern matrix into an enhanced spatiotemporal feature representation.

[0131] In this embodiment, the efficient extraction and interpretation of complex dynamic patterns in switch cabinet helium leak detection data are achieved. The application of adaptive wavelet transform and Wigner-Ville distribution provides a multi-angle analysis of the time-frequency characteristics of the signal. The introduction of persistent homology and algebraic topological features enhances the sensitivity of the method to changes in data structure. The combination of sequence pattern mining and anomaly detection achieves accurate distinction between normal and abnormal patterns. The comprehensive application of these technologies enables the method to extract rich, multi-scale dynamic patterns from the spatiotemporal data of switch cabinet helium leak detection. In particular, the method exhibits excellent recognition capabilities when dealing with complex multi-point leaks or intermittent leak scenarios. Through multi-resolution time-frequency analysis, fast transient leaks and slowly evolving sealing performance degradation can be captured simultaneously. The calculation of topological features helps to identify stable structures in the data, which is particularly important for distinguishing persistent leaks from temporary disturbances. Pattern evolution analysis can not only track the development process of the leak, but also predict its future trend. The introduction of deep time series models further enhances the predictive ability of the method and provides strong support for preventive maintenance. In addition, the application of knowledge graphs and natural language generation technology improves the interpretability of model output, allowing complex data analysis results to be intuitively understood and applied by maintenance personnel.

[0132] According to one aspect of the present application, step S42 is specifically:

[0133] Step S421, read the multi-layer graph structure. Perform spectral decomposition on the graph of each time step. Calculate the eigenvalue distribution of the Laplace matrix. Use the adaptive kernel function to smooth the time series graph structure. Generate a preprocessed time-varying graph sequence.

[0134] Step S422: Read the preprocessed time-varying graph sequence and node attribute set. Design a multimodal attention mechanism to adaptively fuse different types of node features. Use a cross-modal autoencoder to learn the correlation between features. Apply contrastive learning to improve the discriminability of feature representation. Generate a fusion feature matrix.

[0135] Step S423, read the fusion feature matrix. Use the t-SNE algorithm for initial low-dimensional mapping. Apply UMAP to maintain the global structure of the data. Combine local linear embedding (LLE) to capture local features. Use an integration strategy to merge multiple initialization results. Generate an initial dynamic embedding matrix.

[0136] Step S424, read the initial dynamic embedding matrix and the preprocessed time-varying graph sequence. Design a time series smoothing term based on Wasserstein distance. Introduce dynamic time warping (DTW) to measure the similarity of embedded sequences. Construct a time series regularization term to balance static structure preservation and dynamic evolution capture. Generate a time series consistency constraint function.

[0137] Step S425: read the fused feature matrix and the preprocessed time-varying graph sequence. Construct a multi-scale graph convolutional network to capture structural information of different ranges. Design an adaptive pooling layer to dynamically adjust the receptive field size. Introduce skip connections to fuse features of different scales. Generate a multi-scale structural representation.

[0138] Step S426: Read the preprocessed time-varying graph sequence. Design an importance sampling method based on node centrality. Introduce a time-series-aware negative sample generation strategy. Use reinforcement learning to optimize the sampling strategy to maximize the discriminability of the embedding. Generate a dynamic negative sample set.

[0139] Step S427: Read the temporal consistency constraint function, multi-scale structure representation, and dynamic negative sample set. Design an adaptive weight mechanism to dynamically balance different target items. Introduce a curriculum learning strategy to gradually increase the difficulty of the task. Construct a comprehensive optimization objective function, including structure preservation items, temporal consistency items, and contrastive learning items. Generate an optimization objective function.

[0140] Step S428, read the optimization objective function and the initial dynamic embedding matrix. Design an adaptive learning rate adjustment strategy based on Nesterov momentum. Introduce gradient accumulation technology to process large-scale graph data. Use gradient clipping to prevent gradient explosion. Apply second-order optimization methods (such as L-BFGS) to accelerate convergence. Generate an optimized dynamic embedding matrix.

[0141] Step S429: read the optimized dynamic embedding matrix. Use Procrustes analysis to align the embedding spaces at different time steps. Apply the equivariant graph attention network to enhance the rotation invariance of the embedding. Introduce the manifold alignment technology to maintain temporal consistency. Generate a calibrated dynamic embedding matrix.

[0142] Step S4210: Read the calibrated dynamic embedding matrix and the new graph data passed in in real time. Design an incremental update strategy based on a sliding window. Use spectral sparse regularization to maintain the sparsity of the embedding. Apply an online learning algorithm, such as online stochastic gradient descent, to update the embedding in real time. Generate an updated dynamic embedding matrix.

[0143] Step S4211: Read the updated dynamic embedding matrix and the preprocessed time-varying graph sequence. Calculate the link prediction accuracy, node classification performance, and clustering effect. Evaluate the stability and temporal consistency of the embedding. Use persistent homology to analyze the topological characteristics of the embedding. Generate an embedding quality assessment report.

[0144] Step S4212: Read the embedding quality assessment report and the updated dynamic embedding matrix. Based on the assessment results, adaptively adjust the weight of the optimization target. Automatically adjust the hyperparameters using Bayesian optimization. Apply meta-learning strategies to improve the model's adaptability to new tasks. Generate a fine-tuned dynamic embedding matrix.

[0145] Step S4213: Read the fine-tuned dynamic embedding matrix, embedding quality assessment report and multi-layer graph structure. Integrate all analysis results to generate the final dynamic graph embedding report. Associate the dynamic embedding matrix with the original graph structure information to form an enhanced graph representation.

[0146] In this embodiment, efficient dynamic graph embedding of switch cabinet helium leak detection data is achieved. The introduction of graph wave decomposition and multi-scale structural representation provides a basis for capturing topological structures of different scales. The application of adaptive attention mechanism and contrastive learning enhances the discriminative and expressive power of embedding. The design of temporal consistency constraints and incremental update strategies ensures the continuity and real-time performance of embedding in the time dimension. The synergy of these technologies enables the method to transform complex switch cabinet helium leak detection data into low-dimensional, information-rich embedding representations. In particular, when dealing with large-scale, multi-sensor detection systems, the method exhibits excellent computational efficiency and scalability. Through multi-scale graph representation, local micro-leaks and global system state changes can be captured simultaneously. The introduction of the attention mechanism enables the model to automatically focus on the most relevant features and relationships, improving the accuracy of anomaly detection. The combination of contrastive learning and temporal consistency constraints not only enhances the discriminative power of embedding, but also ensures the coherence of temporal analysis. The application of temporal analysis and equivariant graph attention networks solves the problem of rotational invariance of the embedding space, making the embedding results at different time points comparable. In addition, the design of incremental update and online learning strategies enables the method to adapt to the dynamic changes in the operating status of the switchgear, providing the possibility for real-time monitoring and early warning. The introduction of comprehensive performance evaluation and adaptive adjustment mechanism further ensures the stability and effectiveness of the method in long-term applications, laying a solid data foundation for the intelligent operation and maintenance of switchgear.

[0147] According to one aspect of the present application, step S43 is specifically:

[0148] Step S431, read node embedding and multi-layer graph structure. Apply graph wave decomposition to obtain subgraphs of multiple scales. Use spectral coarsening algorithm to coarsen the graph. Construct graph pyramid structure. Generate multi-scale graph representation set.

[0149] Step S432, read the multi-scale graph representation set. For each scale graph: calculate the Forman-Ricci curvature of the k-order neighborhood. Estimate the Betti number of the local homology group. Calculate the persistent heat kernel signature of the node. Extract the local graph kernel features. Integrate the multi-scale features to generate a local topological feature matrix.

[0150] Step S433, read the multi-scale graph representation set. For each scale graph: calculate the spectral entropy and Von Neumann entropy. Extract global persistent homology features. Calculate the Ollivier-Ricci curvature of the graph. Estimate the distortion and asymptotic dimension of the graph. Fuse multi-scale features to generate a global topological feature vector.

[0151] Step S434: read the local topological feature matrix and the global topological feature vector. Design a recursive Bayesian filter to track the time evolution of the feature. Use particle filtering to process non-Gaussian noise. Construct a dynamic topological feature state space model. Generate a time-varying topological feature sequence.

[0152] Step S435: Read the time-varying topological feature sequence and node embedding. Design a self-attention mechanism to capture the long-range dependencies between features. Use a cross-modal Transformer to fuse different types of features. Apply contrastive learning to enhance the discriminability of features. Generate a fused feature tensor.

[0153] Step S436, read the fused feature tensor. Use extreme value theory to model the tail of the feature distribution. Apply adaptive kernel density estimation to fit the multi-dimensional feature distribution. Design an online expectation maximization algorithm to dynamically update the distribution parameters. Generate an adaptive anomaly threshold function.

[0154] Step S437, read the fused feature tensor and the adaptive anomaly threshold function. For each node: calculate the local anomaly score using the isolation forest algorithm. Apply the local outlier factor (LOF) method. Build a local One-Class SVM model. Use the autoencoder reconstruction error as the anomaly indicator. Integrate multiple model results to generate a local anomaly score matrix.

[0155] Step S438, read the fused feature tensor and the adaptive anomaly threshold function. At the full graph scale: apply graph neural network to detect abnormal subgraph structure. Use spectral clustering to identify abnormal communities. Calculate global graph shape statistics to detect overall topological changes. Apply dynamic graph convolutional network to capture time series abnormal patterns. Integrate the results of multiple scales to generate a global abnormal pattern report.

[0156] Step S439: Read the local anomaly scoring matrix and the global anomaly pattern report. Build a graph causal model to infer the propagation path of the anomaly. Use counterfactual reasoning to assess the impact of the anomaly. Apply intervention theory to identify key anomaly sources. Generate an anomaly causal relationship graph.

[0157] Step S4310: read the abnormal causal relationship graph and the time-varying topological feature sequence. Apply the spatiotemporal LSTM network to model the spatiotemporal dependency of the abnormalities. Use the graph attention mechanism to capture the long-range correlation between abnormalities. Construct an abnormal co-occurrence network to analyze the synergistic effect of abnormal patterns. Generate a spatiotemporal abnormal correlation report.

[0158] Step S4311: Read the spatiotemporal anomaly correlation report and anomaly causal relationship graph. Design an explanation generation system based on the knowledge graph. Use SHAP values ​​to quantify the contribution of features to anomalies. Apply a natural language generation model to create human-understandable anomaly descriptions. Generate anomaly explanation reports.

[0159] Step S4312: Read the anomaly explanation report and fused feature tensor. Design a reinforcement learning framework to optimize the anomaly detection strategy. Use active learning methods to select the most informative samples for manual labeling. Apply transfer learning to improve the generalization ability of the model in new scenarios. Update the detection model parameters to generate an optimized anomaly detection model.

[0160] Step S4313: Read the optimized anomaly detection model, anomaly explanation report, and spatiotemporal anomaly correlation report. Calculate multiple evaluation indicators, including precision, recall, F1 score, and AUC. Evaluate the stability and computational efficiency of the model. Generate a comprehensive performance evaluation report. Integrate the multi-layer graph structure, anomaly detection results, and evaluation report to form the final topological anomaly detection result. Output this comprehensive result as the final output of multi-scale adaptive topological anomaly detection.

[0161] In this embodiment, multi-scale adaptive topological anomaly detection of switch cabinet helium leak detection data is implemented. The calculation of Forman-Ricci curvature and persistent homology provides in-depth insights into the topological structure of the data. The application of self-attention mechanism and cross-modal Transformer enhances the effect of feature fusion. The combination of extreme value theory and adaptive kernel density estimation provides a theoretical basis for the dynamic adjustment of anomaly thresholds. The comprehensive application of these technologies enables the method to accurately identify anomalies of various types and scales from complex switch cabinet helium leak detection data. Especially when dealing with large switch cabinet systems or complex multi-point leakage scenarios, the method exhibits excellent detection performance and interpretability. Through multi-scale topological feature analysis, local micro-leakage and global system anomalies can be captured at the same time. The introduction of graph neural network enables the method to make full use of the structural information inside the switch cabinet and improve the accuracy of anomaly location. The application of causal model and counterfactual reasoning can not only identify anomalies, but also infer the propagation path and impact range of anomalies, providing an important basis for fault diagnosis and risk assessment. The combination of spatiotemporal LSTM and graph attention mechanism enhances the method's ability to detect long-term and complex anomaly patterns. In addition, the introduction of knowledge graphs and natural language generation technology improves the interpretability of anomaly detection results, allowing complex data analysis results to be intuitively understood and applied by maintenance personnel. The application of reinforcement learning and active learning strategies enables the system to continuously optimize and self-improve to adapt to the dynamic changes in the switchgear operating environment. The design of comprehensive performance evaluation and model update mechanisms further ensures the effectiveness and reliability of the method in long-term applications, providing strong technical support for preventive maintenance and intelligent operation of switchgear.

[0162] According to one aspect of the present application, step S51 is specifically:

[0163] Step S511, read the graph topology anomaly detection results and the key parameter set used in all previous steps. Construct the objective function set, including detection accuracy, computational efficiency, and model complexity. Define the decision variable space and constraints. Generate the initial population using Latin hypercube sampling. Calculate the objective function value of the initial population. Generate the initial population matrix and the target value matrix.

[0164] Step S512, read the initial population matrix. Dynamically adjust the crossover probability based on population diversity. Design a hybrid strategy of simulated binary crossover (SBX) and differential evolution (DE). Introduce an adaptive crossover distribution index, which is automatically adjusted according to the degree of convergence. Construct an adaptive selection probability distribution for the crossover operation. Generate an adaptive crossover operator.

[0165] Step S513, read the target value matrix. Design an intelligent mutation operator based on Gaussian process. Construct an adaptive adjustment mechanism for mutation intensity, and dynamically adjust it according to the position of the individual on the non-dominated frontier. Introduce Cauchy distribution mutation to enhance the ability to jump out of local optimality. Design a mutation direction guidance strategy and use target gradient information. Generate an intelligent mutation operator.

[0166] Step S514, read the target value matrix of the current population. Implement a fast non-dominated sorting algorithm based on bitmap. Use parallel computing technology to accelerate the sorting process. Introduce an approximate non-dominated sorting strategy to handle large-scale populations. Calculate the dominance number and dominated set of each solution. Generate a non-dominated rank matrix and frontier set.

[0167] Step S515, read the non-dominated rank matrix and the frontier set. Use the K-means++ algorithm to generate the initial reference points in the target space. Design a dynamic reference point adjustment strategy to adaptively update according to the current frontier distribution. Introduce local density estimation to optimize the distribution of reference points in sparse areas. Construct a reference point adaptive update mechanism. Generate an adaptive reference point set.

[0168] Step S516, read the frontier set and the adaptive reference point set. Design a local crowding calculation method based on the reference point. Introduce L∞-norm instead of Euclidean distance to improve calculation efficiency. Use kd tree to speed up the nearest neighbor search. Consider the different scales of the target space and introduce an adaptive normalization strategy. Generate an improved crowding matrix.

[0169] Step S517, read the non-dominated rank matrix, the improved crowding matrix and the current population. Implement the ε-dominance strategy to balance convergence and diversity. Design an elite individual selection mechanism based on reference points. Construct an adaptive pressure regulation strategy for environmental selection to dynamically balance exploration and utilization. Generate a new generation of population matrix.

[0170] Step S518: Read the new generation population matrix and the optimal solution set of previous generations. Calculate the hypervolume index (Hypervolume) and generational distance (Generational Distance). Design an entropy-based convergence evaluation standard. Construct a multi-criteria termination judgment mechanism, comprehensively considering the number of iterations, calculation time and convergence quality. Generate termination condition evaluation results.

[0171] Step S519: Read the termination condition evaluation results and the new generation population matrix. Dynamically adjust the computing resource allocation based on the convergence speed of each target. Design a target switching strategy to focus on different targets at different iteration stages. Build a resource allocation strategy based on reinforcement learning to maximize the overall optimization efficiency. Generate a resource allocation plan.

[0172] Step S5110, read the new generation population matrix and resource allocation scheme. Perform local search on the selected elite individuals. Use a hybrid strategy of pattern search and Nelder-Mead simplex method. Design an adaptive step size adjustment mechanism. Repair solutions that violate constraints and use Lagrange multiplier method to handle constraints. Generate an optimized population matrix.

[0173] Step S5111, read the optimized population matrix. Divide the total population into multiple sub-populations. Design a parallel evolution strategy based on the island model. Construct the migration topology and migration strategy between sub-populations. Introduce a difference evaluation mechanism to maintain the diversity of sub-populations. Design a dynamic population size adjustment strategy. Generate a multi-population matrix after co-evolution.

[0174] Step S5112: read the multi-population matrix after the co-evolution. Use SPEA2 (improved strength Pareto evolutionary algorithm 2) to reconstruct the Pareto frontier. Apply Gaussian process regression to smooth and interpolate the discrete Pareto frontier. Construct a parameterized representation of the frontier to facilitate subsequent decision-making. Generate a reconstructed and smoothed Pareto frontier.

[0175] Step S5113, read the reconstructed and smoothed Pareto frontier. Implement an interactive multi-criteria decision support system. Use the TOPSIS method to sort the solutions. Introduce fuzzy preference relations to deal with the decision maker's uncertain preferences. Construct a solution selection strategy based on regret theory. Generate the final non-dominated solution set and decision recommendation report.

[0176] Step S5114, read the final non-dominated solution set and decision recommendation report. Calculate the performance indicators of multi-objective optimization, including convergence, diversity and stability indicators. Evaluate the computational efficiency and scalability of the algorithm. Generate a comprehensive performance evaluation report. Integrate the non-dominated solution set, decision recommendation and performance evaluation to form the final multi-objective optimization result. Output this comprehensive result as the final output of the adaptive multi-objective collaborative optimization framework.

[0177] In this embodiment, adaptive multi-objective collaborative optimization of the switch cabinet helium leak detection method is realized. The introduction of Latin hypercube sampling and intelligent mutation operators enhances the diversity and search efficiency of the initial population. The application of dynamic reference point adjustment and ε-dominance strategy ensures the convergence and diversity of the optimization process. The combination of parallel evolution and Gaussian process regression further improves the computational efficiency and prediction ability of the algorithm. The synergy of these technologies enables the method to find the best balance between multiple objectives such as detection accuracy, computational efficiency and model complexity. Especially when dealing with large-scale, multi-variable switch cabinet detection systems, the method exhibits excellent optimization performance and adaptability. Through multi-objective optimization, multiple key indicators such as detection sensitivity, false positive rate, and calculation time can be considered at the same time to provide customized solutions for different application scenarios. The design of adaptive crossover and mutation strategies enables the algorithm to adjust the search direction and intensity according to the real-time feedback of the optimization process, effectively avoiding local optimality. The dynamic adjustment of the reference point and the improvement of the congestion calculation ensure the uniform distribution of solutions in the target space and provide a wealth of alternative solutions. The introduction of parallel computing and resource allocation optimization improves the efficiency of the algorithm when dealing with large-scale problems. In addition, the design of the interactive multi-criteria decision support system provides decision makers with a tool to flexibly select the final solution and make trade-offs based on actual needs. The smoothing and interpolation of the Pareto frontier by Gaussian process regression provides a theoretical basis for continuous decision-making. The generation of comprehensive performance evaluation and decision recommendation reports further enhances the usability and credibility of the method, providing comprehensive and scientific decision support for parameter optimization and performance tuning of the switch cabinet helium leak detection system.

[0178] In another embodiment of the present application, the specific implementation process is as follows:

[0179] Step S11, obtain data from at least two sensors (different switch cabinets may belong to different manufacturers and be equipped with different sensors), and obtain helium concentration data a(t) from the helium mass spectrometer. Obtain pressure distribution data b(t) from the pressure sensor array. Obtain temperature distribution data c(t) from the temperature sensor network. Obtain acoustic signal data d(t) from the acoustic sensor array. Obtain electromagnetic field intensity data e(t) from the electromagnetic field sensor. Calculate the rate of change of each sensor signal v_i(t)=|d_i(t) / dt|, where i represents the sensor type. Based on the rate of change v_i(t), calculate the adaptive sampling frequency f(t)=max(f_min, min(f_max, k*max(v_i(t)))), where f_min and f_max are the minimum and maximum sampling frequencies, respectively, and k is the proportional coefficient. Use the calculated f(t) to resample all sensor data. Organize the resampled data into the original data tensor R(t, f(t), [a, b, c, d, e]).

[0180] Step S12, read the original data tensor R(t, f(t), [a, b, c, d, e]) obtained in step S11. For each sensor signal i in R, apply the adaptive multi-resolution decomposition algorithm. First, use wavelet transform to perform multi-scale decomposition on the signal to obtain the initial decomposition set {W_i}_j, where j represents the decomposition level. Then, apply empirical mode decomposition to each {W_i}_j to obtain the multi-scale component set {IMF_i}_j. Calculate the information entropy H_j and computational complexity C_j of each decomposition level j. Define the objective function O_j=w1*H_j-w2*C_j, where w1 and w2 are weight coefficients. Select the level j_opt that maximizes O_j as the optimal decomposition level. Based on j_opt, select the corresponding multi-scale component from {IMF_i}_j. Combine all processed signals to generate a six-dimensional decomposition data tensor D(t, f(t), [a, b, c, d, e], j).

[0181] Step S13, read the decomposed data tensor D(t, f(t), [a, b, c, d, e], j) obtained in step S12. For each decomposition level j of each sensor signal i in D, construct a time delay embedding vector Vi_ij(t)=[x_ij(t), x_ij(t+τ), ..., x_ij(t+(m-1)τ)], where τ is the time delay and m is the embedding dimension. Construct a Vietoris-Rips complex based on Vi_ij(t) and calculate 0-dimensional and 1-dimensional persistence homology. Obtain a persistence graph PD_ij, which includes birth time α_k and death time β_k. Calculate the anomaly measure AD_ij=Σ_k(β_k-α_k) / (t_max-t_min). Compare AD_ij with a pre-set adaptive threshold τ_ij. If AD_ij>τ_ij, mark the data point as abnormal. For data points marked as abnormal, use the local linear regression method to repair them. The repaired data is reorganized to generate a cleaned six-dimensional data tensor C(t, f(t), [a, b, c, d, e], j).

[0182] Step S14, read the cleaned data tensor C(t, f(t), [a, b, c, d, e], j) obtained in step S13. For each decomposition level j of each sensor signal i in the cleaned data tensor C, use the Isomap algorithm to embed it into a high-dimensional manifold to obtain the manifold M_ij. Calculate the geodesic distance d_g(M_ij, M_kl)=min(Σ_pd_E(x_p, x_{p+1})) between manifolds, where d_E is the Euclidean distance and x_p is the point on the geodesic. Set the sensor importance weight w_ij and construct the objective function E=Σ_i, j, k, lw_ij*w_kl*d_g(M_ij, M_kl). Use the gradient descent method to minimize the objective function E and obtain the optimal alignment parameter θ*. Based on θ*, the data in the cleaned data tensor C is transformed to obtain the aligned six-dimensional data tensor A(t, f(t), [a, b, c, d, e], j).

[0183] Step S15, read the aligned data tensor A(t, f(t), [a, b, c, d, e], j) obtained in step S14. Apply short-time Fourier transform to each decomposition level j of each sensor signal i in A to obtain the time-frequency representation S_ij(t, ω). Calculate the statistical moments of S_ij(t, ω), including mean μ_ij(t), variance σ²_ij(t), skewness γ_ij(t) and kurtosis κ_ij(t). Extract geometric features, including fractal dimension D_ij(t) and Lyapunov exponent λ_ij(t). Combine all features to form a feature set F_ij(t)=[μ_ij(t), σ²_ij(t), γ_ij(t), κ_ij(t), D_ij(t), λ_ij(t)]. Calculate the Shannon entropy H(F_k) and mutual information between features I(F_k; F_l) for each feature. Define feature importance score S(F_k)=H(F_k)-Σ_l≠kI(F_k;F_l). Sort features based on S(F_k) and select the n features with the highest scores. Organize the selected features into a dynamic feature matrix F(t, [f_1(t), ..., f_n(t)]).

[0184] Step S21, read the aligned data tensor A(t, f(t), [a, b, c, d, e], j) output by step S1. Initialize Tucker rank (R1, R2, R3, R4, R5, R6) and CP rank R. Apply Tucker decomposition to A to obtain the core tensor G and factor matrix U_i. Calculate the Tucker reconstruction error E_T=||AG×1U_1×2U_2×3U_3×4U_4×5U_5×6U_6||_F / ||A||_F. Apply CP decomposition to A to obtain the factor matrix V_i. Calculate the CP reconstruction error E_C. Compare E_T and E_C, and select the decomposition method with smaller error. If Tucker decomposition is selected, L=G×1U_1×2U_2×3U_3×4U_4×5U_5×6U_6; if CP decomposition is selected, L=Σ_r=1 R v_1 r ○v_2 r ○v_3 r ○v_4 r ○v_5 r ○v_6 r , where ○ represents the outer product and r represents the index variable. Reshape L into a low-dimensional representation L(t, r, [a, b, c, d, e]).

[0185] Step S22, read the low-dimensional representation L(t, r, [a, b, c, d, e]) obtained in step S21. Define the time scale set S={s_1, s_2, ..., s_K}. For each time scale s_k and each pair of sensors (i, j), calculate the multi-scale Granger causality. First, perform a wavelet transform on L at scale s_k to obtain the wavelet coefficients W_i(s_k, t) and W_j(s_k, t). Construct a VAR model: W_i(s_k, t)=Σ_p=1 P a_pW_i(s_k,tp)+Σ_p=1 P b_pW_j(s_k,tp)+ε(t). Calculate the F statistic F_ij(s_k)=(RSS_R-RSS_U) / RSS_U*(T-2P-1) / P, where RSS_R and RSS_U are the residual sums of squares of the restricted and unrestricted models respectively, T is the number of samples, and P is the lag order. Calculate the multiscale causal intensity MCI(i,j,s_k)=Σ_tw(s_k,t)*F_ij(s_k,t), where w(s_k,t) is the time-varying weight function. Organize all MCI(i,j,s_k) into a multiscale causal adjacency matrix C(s,t,[a,b,c,d,e]).

[0186] Step S23, read the multi-scale causal adjacency matrix C(s, t, [a, b, c, d, e]) obtained in step S22 and the dynamic feature matrix F(t, [f_1(t), ..., f_n(t)] output in step S1. For each pair of sensors (i, j) and each time point t, calculate the conditional entropy H(X_i(t)|X_j(t))=-Σ_x_i, x_jp(x_i, x_j)log(p(x_i|x_j)), where the probability is estimated from F using the kernel density estimation method. Calculate the information flow IF(i, j, t)=Σ_sMCI(i, j, s, t)*H(X_i(t)|X_j(t)). Construct a time-varying information flow network G(t), where the nodes represent sensors and the edge weights are IF(i, j, t). Apply a community detection algorithm, such as the Louvain method, to G(t) to identify the dynamic community structure. Calculate the centrality indicators of each node, such as degree centrality and eigenvector centrality. Represent G(t) as a combination of the adjacency matrix and the node attribute matrix G(t, [a, b, c, d, e]). Finally, combine the low-dimensional representation L, the multi-scale causal adjacency matrix C, and the time-varying information flow network G into a fused multimodal representation M(t, [L, C, G]).

[0187] Step S31, read the low-dimensional representation L(t, r, [a, b, c, d, e]) in the fused multimodal representation M(t, [L, C, G]) output by step S2. For each sensor component in L, calculate the mutual information function MI(τ)=Σp(x(t), x(t+τ))log(p(x(t), x(t+τ)) / (p(x(t))p(x(t+τ)))), where the probability is calculated using the kernel density estimation method. Select τ that makes MI(τ) reach the local minimum for the first time as the optimal time delay τ_opt. Use the error prediction method to estimate the optimal embedding dimension m_opt. Construct a candidate embedding vector Y_m(t)=[x(t), x(t+τ_opt), ..., x(t+(m-1)τ_opt)], where m increases from 1. For each m, use the k nearest neighbor method to predict x(t+T) and calculate the prediction error E(m). When E(m) no longer decreases significantly, the value of m is taken as m_opt. Use τ_opt and m_opt to construct the phase space trajectory X(t, τ_opt, m_opt) = [x(t), x(t+τ_opt), ..., x(t+(m_opt-1)τ_opt)]. Repeat this process for all sensor components to obtain the complete phase space trajectory X(t, τ, m).

[0188] Step S32, read the phase space trajectory X(t, τ, m) obtained in step S31. Apply a sliding window to X with a window size of w and a step size of s to obtain a time series {X_i}, where i represents the window index. Construct a Vietoris-Rips complex VR_i(ε) for each X_i, where ε is the distance threshold. Calculate the k-dimensional persistence homology (k=0, 1, 2) of VR_i(ε) to obtain the persistence graph PD_k,i. Combine the persistence graphs of all windows into a time-varying persistence graph PD_k(t). Calculate the statistical features of PD_k(t), including the persistence entropy PE_k(t)=-Σ(l_j / Σl_j)log(l_j / Σl_j), where l_j is the length of the j-th persistence interval, and the Betti number curve β_k(t, ε). Define the topological complexity metric TC(t)=Σ_kw_k*(PE_k(t)+Σ_εβ_k(t,ε) / max(β_k(t,ε))), where w_k is the weight coefficient. Combine PD_k(t) and TC(t) into a topological feature representation TF(t, [PD, TC]).

[0189] Step S33: read the phase space trajectory X(t, τ, m) obtained in step S31 and the topological feature representation TF(t, [PD, TC]) obtained in step S32. Apply continuous wavelet transform CWT(t, s) = ∫X(t')ψ*((t'-t) / s)dt' to X, where ψ is the wavelet function and s is the scale parameter. Calculate the wavelet energy spectrum WE(t, s) = |CWT(t, s)| 2 . Identify the significant peaks in WE(t, s) and obtain the candidate pattern set {C_i(t, s)}. For each C_i(t, s), calculate its duration D_i and energy E_i. Define the pattern importance index PI_i=E_i*D_i / Σ_j(E_j*D_j). Sort the candidate patterns according to PI_i and select the top p patterns as the main dynamic patterns. For each selected pattern, extract its time-frequency feature TF_i and shape feature SF_i. Combine TF_i and SF_i with the topological feature TF of the corresponding time period to form a pattern descriptor PD_i=[TF_i, SF_i, TF]. Organize all PD_i into a pattern matrix P(t, p). Finally, the phase space trajectory X, the persistence diagram PD_k and the pattern matrix P are combined into a spatiotemporal feature representation TS(t, [X, PD, P]).

[0190] Step S41, read the fused multimodal representation M(t, [L, C, G]) output by step S2 and the spatiotemporal feature representation TS(t, [X, PD, P]) output by step S3. Define a layer set Λ={λ_1, λ_2, ..., λ_K}, where each layer represents a different type or scale of interaction. For each layer λ_k, construct a node set V_k, including sensor nodes and virtual nodes generated by dynamic patterns in TS. Construct an edge set E_k based on similarities in L, causal relationships in C, and information flow in G. Define a node attribute set A_k, including features in L, dynamic features in X, and topological features in PD. Calculate the inter-layer coupling strength IC(λ_i, λ_j)=Σ_v, wE(v_i, w_j) / (|V_i|*|V_j|), where E(v_i, w_j) is the weight of the cross-layer edge. Construct the inter-layer edge set E_ij to connect nodes in different layers that represent the same entity or are highly related. Integrate the information of all layers to form a multi-layer graph structure ML-G(t, [V, E, A]), where V=∪V_k, E=(∪E_k)∪(∪E_ij), A=∪A_k.

[0191] Step S42: read the multi-layer graph structure ML-G(t, [V, E, A]) constructed in step S41. Initialize the node embedding matrix Z(0)∈R (|V|×d) , d is the embedding dimension. Define the intra-layer loss function L_intra=Σ_kΣ_(i, j∈V_k)||Z_i-Z_j|| 2 *A_k(i, j), where A_k is the adjacency matrix of the kth layer. Define the inter-layer loss function L_inter=Σ_(i, j∈V, l(i)≠l(j))||Z_i-Z_j|| 2 *IC(l(i), l(j)), where l(i) represents the layer to which node i belongs. Define the attribute preservation loss function L_attr=Σ_i||Z_i-f(A_i)|| 2 , where f is the attribute conversion function. Construct the overall objective function L=α*L_intra+β*L_inter+γ*L_attr, where α, β, and γ are trade-off coefficients. Use stochastic gradient descent to optimize L, and the update rule is Z(t+1)=Z(t)-η▽L, where η is the learning rate and ▽ represents the gradient. Stop the optimization when L converges or reaches the maximum number of iterations. Output the final node embedding Z(t, d).

[0192] Step S43, read the node embedding Z(t, d) obtained in step S42 and the multi-layer graph structure ML-G(t, [V, E, A]) constructed in step S41. For each node v, calculate the local topological feature LTF(v, t). First, extract the k-hop neighborhood N_k(v) of v. Calculate the local clustering coefficient LCC(v)=2|E(N_k(v))| / (|N_k(v)|*(|N_k(v)|-1)). Calculate the local heterogeneity coefficient LAC(v)=Σ_(u∈N_k(v))|deg(v)-deg(u)| / (|N_k(v)|*deg(v)). Calculate the local page rank LPR(v), and use the PersonalRank algorithm to iteratively calculate on N_k(v). Combining these features, LTF(v, t)=[LCC(v), LAC(v), LPR(v)]. Calculate the global topological feature GTF(v, t). Apply the spectral clustering algorithm to cluster Z(t, d) and obtain the community affiliation C(v) of the node. Calculate the centrality indicators of node v, including the eigenvector centrality EVC(v) and the betweenness centrality BC(v). Calculate the structural role SR(v) of node v, using the RolX algorithm based on the local and global feature extraction of the node. Combining these features to obtain GTF(v, t) = [C(v), EVC(v), BC(v), SR(v)]. Define the anomaly score AS(v, t) = w_1*D_L(LTF(v, t), μ_L)+w_2*D_G(GTF(v, t), μ_G), where D_L and D_G are the distance functions of local and global features, μ_L and μ_G are the corresponding reference distributions, and w_1 and w_2 are weight coefficients. ML-G, Z and AS are combined into the graph topology anomaly detection result GT(t, [ML-G, Z, AS]).

[0193] Step S51, read the graph topology anomaly detection result GT(t, [ML-G, Z, AS]) output by step S4 and the key parameter set P={p_1, p_2, ..., p_n} used in all previous steps. Define the detection accuracy objective function f_1(P)=1-AUC(ROC(AS)), where ROC is the receiver operating characteristic curve and AUC is the area under the curve. Define the computational efficiency objective function f_2(P)=T_compute(P) / T_max, where T_compute is the total computing time and T_max is the maximum allowed computing time. Construct a constraint set C={c_1(P)≤0, c_2(P)≤0, ..., c_m(P)≤0}, including parameter value ranges and resource constraints. Initialize the population X={x_1, x_2, ..., x_N}, where each individual x_i represents a set of parameter configurations. For each x_i, calculate the objective function values ​​f_1(x_i) and f_2(x_i). Use non-dominated sorting to stratify the population and obtain the non-dominated level R(x_i). Calculate the crowding distance D(x_i)=Σ_k(f_k(x_i+1)-f_k(x_i-1)) / (f_k_max-f_k_min). Perform selection, crossover and mutation operations based on R(x_i) and D(x_i) to generate a new population. Repeat the steps from initializing the population to generating a new population until the maximum number of iterations is reached or convergence occurs. Output the final non-dominated solution set NS(P_opt).

[0194] Step S52, read the non-dominated solution set NS(P_opt) obtained in step S51 and the graph topology anomaly detection result GT(t, [ML-G, Z, AS]) output by step S4. Define the performance index PI(t)=w_1*Precision(t)+w_2*Recall(t)-w_3*FPR(t), where Precision, Recall and FPR are precision, recall and false positive rate respectively, and w_1, w_2 and w_3 are weight coefficients. Initialize the Q function Q(s, a)=0, where s is the state, indicating the current parameter configuration, and a is the action, indicating the parameter adjustment strategy. Define the state transition function T(s, a, s')=P(s'|s, a), which indicates the probability of transferring from state s to state s' after executing action a. Define the reward function R(s, a)= ΔPI(t), which is the change in the performance index. Use the Q-learning algorithm to update the Q function: Q(s, a)←Q(s, a)+α[R(s, a)+γmax_a'Q(s', a')-Q(s, a)], where α is the learning rate and γ is the discount factor. At each time step t, select action a_t with the ε-greedy strategy, select max_aQ(s_t, a) with probability 1-ε, and select randomly with probability ε. Execute the selected action a_t, update the parameter configuration, and get the new state s_t+1. Observe the reward R(s_t, a_t) and update the Q function. Based on the updated Q function, generate the time-varying optimal parameter set P_opt(t). Combine GT and P_opt(t) into the optimized anomaly detection result OD(t, [GT, P_opt]).

[0195] It should be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations.

Claims

1. Helium leak detection method for switch cabinets, It is characterized in that The steps include: Step S1, acquiring raw data from the sensor in real time, organizing the raw data into an N-dimensional raw data tensor, applying an adaptive multi-resolution decomposition algorithm to the raw data tensor, and obtaining a decomposed data tensor; Based on the decomposed data tensor, a persistence graph is constructed and persistence coherence is calculated to detect and repair anomalies and generate a cleaned data tensor. Perform nonlinear manifold alignment on the cleaned data tensor to obtain an aligned data tensor, extract dynamic features from the aligned data tensor, and generate a dynamic feature matrix; N is a natural number greater than 1; Step S2, obtaining the aligned data tensor and applying an adaptive tensor decomposition algorithm to it to obtain a low-dimensional representation; Construct a multi-scale causal network based on low-dimensional representation and generate a multi-scale causal adjacency matrix; construct a time-varying information flow network using the multi-scale causal adjacency matrix and the dynamic feature matrix; construct a fused multimodal representation based on low-dimensional representation, multi-scale causal adjacency matrix and time-varying information flow network; Step S3, reading the fused multimodal representation, reconstructing the low-dimensional representation in the multimodal representation into a nonlinear dynamic system, and obtaining a phase space trajectory; Calculate persistence coherence based on phase space trajectory and generate persistence graph; extract dynamic patterns from phase space trajectory and persistence graph to obtain pattern matrix; Combine phase space trajectories, persistence maps, and pattern matrices into a spatiotemporal feature representation; Step S4: read and construct a multi-layer graph structure based on the fused multimodal representation and spatiotemporal feature representation, apply a dynamic graph embedding algorithm to the multi-layer graph structure to obtain a low-dimensional representation of the node, calculate local and global topological features based on the low-dimensional representation of the node, and construct anomaly detection indicators; combine the multi-layer graph structure, the low-dimensional representation of the node, and the anomaly detection indicator into a graph topology anomaly detection result; Step S5, reading the graph topology anomaly detection result and constructing a key parameter set and a multi-objective optimization framework to obtain a non-dominated solution set, constructing a dynamic parameter adjustment strategy based on the non-dominated solution set, and generating a time-varying optimal parameter set; The graph topology anomaly detection results and the time-varying optimal parameter set are combined into the optimized anomaly detection results.

2. The helium leak detection method for a switch cabinet according to claim 1, It is characterized in that Step S1 is specifically as follows: Step S11, acquiring part of the raw data from at least two sensors in real time; calculating the rate of change of each sensor signal according to the part of the raw data, and determining the adaptive sampling frequency; resampling all sensor data using the adaptive sampling frequency; and organizing the resampled data into a raw data tensor; Step S12, reading the raw data tensor, applying an adaptive multi-resolution decomposition algorithm to each sensor signal in the tensor, and performing multi-scale decomposition on the signal using wavelet transform to obtain an initial decomposition set; Applying empirical mode decomposition to the initial decomposition set to obtain a multi-scale component set; Calculate the information entropy and computational complexity of each decomposition level; construct and optimize the objective function and select the optimal decomposition level; Based on the optimal level, select the corresponding component from the multi-scale component set; combine the processed signals to generate a decomposed data tensor; Step S13, reading the decomposed data tensor, constructing a time delay embedding vector for each decomposition level of each sensor signal in the decomposed data tensor; constructing a complex based on the embedding vector, calculating the persistence data, and obtaining a persistence graph; calculating anomaly metrics, and comparing the anomaly metrics with a preset threshold, and marking abnormal data points; Use local linear regression method to repair abnormal data points; reorganize the repaired data to generate cleaned data tensors; Step S14, reading the cleaned data tensor and embedding each decomposition level of each sensor signal into a high-dimensional manifold using a dimensionality reduction algorithm; Calculate the geodesic distance between high-dimensional manifolds; set sensor importance weights, construct the objective function, and use the optimization algorithm to minimize the objective function to obtain the optimal alignment parameters; based on the optimal alignment parameters, transform the data to obtain the aligned data tensor; Step S15, reading the aligned data tensor, and applying a time-frequency analysis method to each decomposition level of each sensor signal in the data tensor; Compute statistical moments of time-frequency representation; extract geometric features; Form a feature set; calculate the information entropy of each feature and the mutual information between features; and construct feature importance scores, sort and select features based on the importance scores; organize the selected features into a dynamic feature matrix.

3. The helium leak detection method for a switch cabinet according to claim 1, It is characterized in that Step S2 is specifically as follows: Step S21, read the aligned data tensor and dynamic feature matrix; initialize the decomposition rank parameter; apply Tucker decomposition to the data tensor to obtain a core tensor and a factor matrix; calculate the Tucker reconstruction error; apply CP decomposition to the data tensor to obtain a factor matrix; Calculate the CP reconstruction error, compare the errors of the two decomposition methods, and select the method with the smaller error; reconstruct the data based on the selected decomposition method to obtain a low-dimensional representation; reshape the low-dimensional representation into a low-dimensional data tensor; Step S22, read the low-dimensional data tensor; construct a time scale set; for each pair of sensor signals in the low-dimensional data tensor, calculate the multi-scale Granger causality at each time scale; perform wavelet transform on the low-dimensional data tensor at each scale to obtain wavelet coefficients; construct a vector autoregression model; calculate the F statistic; calculate the multi-scale causal strength; organize all the multi-scale causal strengths into a multi-scale causal adjacency matrix; Step S23, reading a multi-scale causal adjacency matrix and a dynamic feature matrix; For each pair of sensor signals in the dynamic feature matrix, the conditional entropy is calculated at each time point; the probability is estimated from the dynamic feature matrix using the kernel density estimation method; the information flow is calculated; and a time-varying information flow network is constructed with sensors as nodes and information flow as edge weights; Apply community detection algorithms to time-varying information flow networks to identify dynamic community structures; calculate the centrality index of each node in the time-varying information flow network; represent the time-varying information flow network as a combination of an adjacency matrix and a node attribute matrix; combine low-dimensional data tensors, multi-scale causal adjacency matrices, and time-varying information flow networks into a fused multimodal representation.

4. The helium leak detection method for a switch cabinet according to claim 1, It is characterized in that Step S3 is specifically as follows: Step S31, reading the low-dimensional data tensor in the fused multimodal representation; calculating the mutual information function for each sensor component in the low-dimensional data tensor; selecting the time delay that makes the mutual information function reach the local minimum for the first time as the optimal time delay; Use the error prediction method to estimate the optimal embedding dimension; Construct candidate embedding vectors; for each candidate embedding dimension, use the k-nearest neighbor method to predict future values ​​and calculate the prediction error; when the prediction error no longer decreases significantly, use the dimension as the optimal embedding dimension; Construct a phase space trajectory using the optimal time delay and optimal embedding dimension; repeat this process for all sensor components in the low-dimensional data tensor to obtain the complete phase space trajectory; Step S32, read the complete phase space trajectory; apply a sliding window to the phase space trajectory to obtain a time series set; construct a Vietoris-Rips complex for each time series in the time series set; calculate the multidimensional persistence homology of the complex to obtain a persistence graph; combine the persistence graphs of all windows into a time-varying persistence graph; calculate the statistical features of the time-varying persistence graph, including persistence entropy and Betti number curves; define a topological complexity measure; combine the time-varying persistence graph and the topological complexity measure into a topological feature representation; Step S33, reading the complete phase space trajectory and topological feature representation; Applying continuous wavelet transform to the phase space trajectory; calculating the wavelet energy spectrum; identifying significant peaks in the wavelet energy spectrum to obtain a candidate pattern set; For each candidate mode in the candidate mode set, calculate its duration and energy; define a mode importance index; sort the candidate modes according to the mode importance index and select the main dynamic mode; For each selected main dynamic mode, extract its time-frequency characteristics and shape characteristics; The extracted features are combined with the topological features of the corresponding time period to form pattern descriptors; all pattern descriptors are organized into a pattern matrix; and the phase space trajectory, time-varying persistence diagram and pattern matrix are combined into a spatiotemporal feature representation.

5. The helium leak detection method for a switch cabinet according to claim 1, It is characterized in that Step S4 specifically comprises: Step S41, read the fused multimodal representation and spatiotemporal feature representation; define a layer set, each layer represents a different interaction type or scale; for each layer in the layer set, construct a node set, including sensor nodes and virtual nodes generated by dynamic patterns in the spatiotemporal feature representation; construct an edge set based on similarities in low-dimensional data tensors, causal relationships in multi-scale causal adjacency matrices, and information flows in time-varying information flow networks; define a node attribute set, including features in low-dimensional data tensors, dynamic features in phase space trajectories, and topological features in time-varying persistence graphs; Calculate the interlayer coupling strength; Construct inter-layer edge sets to connect nodes in different layers that represent the same entity or are highly related; integrate information from all layers to form a multi-layer graph structure; Step S42, read the multi-layer graph structure; initialize the node embedding matrix; Define the intra-layer loss function based on the adjacency matrix of each layer; define the inter-layer loss function based on the inter-layer coupling strength; define the attribute preservation loss function based on the node attribute set; construct the overall objective function, combining the intra-layer loss, inter-layer loss and attribute preservation loss; use the stochastic gradient descent method to optimize the overall objective function; update the node embedding matrix until the objective function converges or the maximum number of iterations is reached; output the final node embedding; Step S43, reading node embedding and multi-layer graph structure; for each node in the multi-layer graph structure, calculating local topological features; extracting the k-hop neighborhood of the node; calculating the local clustering coefficient; calculating the local heterogeneity coefficient; calculating the local page rank; combining these features to obtain local topological features; Calculate global topological features; Apply spectral clustering algorithm to node embedding to obtain the community affiliation of nodes; calculate the centrality indicators of nodes, including eigenvector centrality and betweenness centrality; calculate the structural role of nodes; combine these features to obtain global topological features; define anomaly scores, combining local and global topological features; combine multi-layer graph structure, node embedding and anomaly scores into graph topology anomaly detection results.

6. The helium leak detection method for a switch cabinet according to claim 1, It is characterized in that Step S5 specifically includes: Step S51, read the graph topology anomaly detection result and the key parameter set used in all previous steps; define the detection accuracy objective function, based on the area under the curve of the receiver operating characteristic curve of the anomaly score; define the computational efficiency objective function, based on the total computing time and the maximum allowed computing time; construct a set of constraints, including parameter value ranges and resource restrictions; initialize the population, each individual represents a set of parameter configurations; for each individual in the population, calculate the objective function value; use non-dominated sorting to stratify the population to obtain a non-dominated level; calculate the crowding distance; perform selection, crossover and mutation operations based on the non-dominated level and crowding distance to generate a new population; Until the maximum number of iterations is reached or convergence occurs; output the final non-dominated solution set; Step S52, read the non-dominated solution set and the graph topology anomaly detection result; define the performance index, combining the precision, recall and false positive rate; initialize the Q function, the state represents the current parameter configuration, and the action represents the parameter adjustment strategy; Define a state transition function, which represents the probability of transitioning from one state to another after executing an action; define a reward function, which is based on the change in performance indicators; Use the Q-learning algorithm to update the Q function; at each time step, select an action using the ε-greedy strategy; execute the selected action, update the parameter configuration, and obtain a new state; Observe the reward and update the Q function; generate a time-varying optimal parameter set based on the updated Q function; combine the graph topology anomaly detection result and the time-varying optimal parameter set into an optimized anomaly detection result.

7. The helium leak detection method for a switch cabinet according to claim 2, It is characterized in that In the step S12, specifically: Step S121, read the original data tensor; apply the sliding window technology to each sensor signal in the data tensor; use the improved box counting method to estimate the local fractal dimension in each window; calculate the fractal spectrum of the signal to obtain the change of the fractal dimension over time; store the fractal dimension time series as a fractal feature matrix; Step S122, reading the fractal feature matrix; Use dynamic programming algorithm to find the optimal segmentation point on the fractal dimension time series; each segment corresponds to a characteristic scale; Calculate the average fractal dimension of each segment; Adaptively select the decomposition basis function family according to the average fractal dimension; Store the selected basis function family and the corresponding scale information as a scale selection matrix; Step S123, read the original data tensor and the scale selection matrix; decompose each sensor signal in the data tensor at each adaptively selected scale; use the basis function family corresponding to the scale to decompose the signal; calculate the decomposition coefficient at each scale; combine the decomposition coefficients of all scales into a multi-scale decomposition tensor; Step S124, reading a multi-scale decomposition tensor; applying kernel principal component analysis to the coefficients of each scale in the decomposition tensor to extract nonlinear features; Use radial basis function kernel for nonlinear mapping; select principal components that can explain 90% of the variance; store the extracted nonlinear features as feature tensors; Step S125, read the feature tensor and the original data tensor; use the particle swarm optimization algorithm to find the optimal reconstruction weight; the objective function is the weighted sum of the reconstruction error and the information entropy; for each time point, based on the current optimal weight, reconstruct the features in the feature tensor into signals; combine the reconstructed signals into a reconstructed data tensor; Step S126, reading the reconstructed data tensor and the original data tensor; calculating the residual between the reconstructed data tensor and the original data tensor; Use adaptive Kalman filter to process the residual; The filter parameters are dynamically adjusted according to the statistical characteristics of the residual; Add the filtered residual to the reconstructed data tensor to obtain a refined reconstructed data tensor; Step S127, reading the finely reconstructed data tensor and the original data tensor; Calculate multiple evaluation metrics, including root mean square error, structural similarity index, and information retention rate; Use fuzzy comprehensive evaluation method to combine multiple indicators into a single quality score; if the quality score is lower than the preset threshold, return to step S122 to adjust parameters and re-decompose; Step S128, reading the fine reconstruction data tensor, the multi-scale decomposition tensor and the feature tensor; Align the refined reconstructed data tensor with the size and structure of the original data tensor; Combining the multi-scale decomposition tensor and the feature tensor to generate an enhanced feature tensor; The refined reconstructed data tensor and the enhanced feature tensor are merged to form the final decomposed data tensor.

8. The helium leak detection method for a switch cabinet according to claim 2, It is characterized in that Step S14 is specifically as follows: Step S141, read the cleaned data tensor; apply the t-SNE algorithm to perform initial dimensionality reduction on each decomposition level of each sensor signal in the data tensor; use cosine similarity as the distance metric; Generate an initial low-dimensional embedding matrix; Step S142, reading an initial low-dimensional embedding matrix; Use dynamic time warping algorithm to calculate the similarity between time series; Based on DTW distance, an adaptive Gaussian kernel function is constructed; Calculate the kernel matrix; Store the kernel matrix as a similarity matrix; Step S143, reading the similarity matrix and the initial low-dimensional embedding matrix; Construct the graph Laplacian matrix; define the manifold consistency objective function, combining local preservation and global alignment; optimize the objective function using the alternating direction multiplier method; Update the low-dimensional embedding matrix; store the optimized low-dimensional embedding matrix as the aligned embedding matrix; Step S144, reading the alignment embedding matrix; Use recurrent neural network RNN ​​to learn time-varying mapping functions; the input of RNN is the embedding vector of adjacent time steps, and the output is the predicted embedding of the next time step; minimize the prediction error and manifold consistency loss; store the learned RNN model parameters as a time-varying mapping model; Step S145, read the time-varying mapping model and the alignment embedding matrix; use the learned RNN model to interpolate or extrapolate the data at missing or future time points; generate a complete time series embedding; merge the interpolation and extrapolation results with the original alignment embedding matrix to form an extended embedding matrix; Step S146, reading the extended embedding matrix; Decompose the embedding into multiple scales using wavelet transform; apply local linear embedding algorithm at each scale; merge the embedding results of different scales by weighted summation; The weights are dynamically adjusted using an adaptive fuzzy inference system; Generate a fused multi-scale embedding matrix; Step S147, reading the fused multi-scale embedding matrix; Constructing continuous homology features; Use persistent homology information to guide manifold deformation; minimize Wasserstein distance to preserve topology; apply discrete exterior differential operators for smoothing; generate topology-preserving embedding matrices; Step S148, reading the topology-preserving embedding matrix and the original data tensor; calculating the geodesic distance preservation rate before and after the manifold alignment; Evaluate the degree of preservation of local neighborhood structure; use the mutual information criterion to quantify the degree of alignment between different sensor signals; combine multiple indicators to generate an alignment quality score; Step S149, reading the topology-preserving embedding matrix, alignment quality score, and original data tensor; Adaptively adjust embedding weights based on alignment quality scores; Map the weighted embedding back to the original data space; fuse it with the original data tensor to generate the final aligned data tensor.

9. The helium leak detection method for a switch cabinet according to claim 3, It is characterized in that Step S21 is specifically as follows: Step S211, read the aligned data tensor and dynamic feature matrix; standardize the data tensor; calculate the modal correlation matrix of the tensor; group the modal using the spectral clustering algorithm; generate the modal grouping information matrix; Step S212, reading the modality grouping information matrix; for each modality group, calling a predetermined kernel function; Automatically select kernel function hyperparameters using Gaussian process regression; Generate a kernel function parameter matrix; apply the kernel function to the original data tensor to generate a kernelized data tensor; Step S213, reading the kernelized data tensor; Apply tensor singular value decomposition to each time window using a sliding window technique; compute singular value decay curves; estimate the optimal rank for each window using the Bayesian Information Criterion; predict the ranks of future time windows using a long short-term memory network; Generate a dynamic rank estimation sequence; Step S214, read the kernelized data tensor and the dynamic rank estimation sequence; apply different tensor decomposition methods to each modality group: use Tucker decomposition for dense modality groups; use CP decomposition for sparse modality groups; use tensor column selection algorithm for mixed modality groups; perform decomposition operations in combination with dynamic rank information; Generates a set of factorized matrices and core tensors; Step S215, read the decomposition factor matrix set; apply structured sparse regularization to each factor matrix; use the proximal gradient descent algorithm to optimize the regularization objective function; introduce a temporal smoothing constraint to ensure the temporal continuity of the factor; generate a regularized factor matrix set; Step S216, read the regularized factor matrix set and core tensor; call the adaptive fusion network based on the attention mechanism; Use reinforcement learning to train the fusion strategy, and the reward function combines the reconstruction error and information retention rate; generate the optimal fusion weight matrix; Step S217, reading the fusion weight matrix and the new data transmitted in real time; Adopt an incremental update rule based on a randomized approximation algorithm; perform online updates on factor matrices and core tensors; use exponential moving averages to maintain long-term statistical information; Generate updated decomposition results; Step S218, reading the updated decomposition result and the original data tensor; calculating the reconstruction error tensor; Perform principal component analysis on the error tensor to extract the main error modes; Call the pre-built wavelet neural network based error predictor; generating an error compensation model; Step S219, reading the updated decomposition result, error compensation model and original data tensor; Reconstruct the data tensor based on the decomposition results to obtain a low-dimensional representation.

10. A helium leak detection system for switch cabinets, It is characterized in that include: at least one processor; as well as, a memory communicatively connected to at least one of the processors; wherein, The memory stores instructions executable by the processor, and the instructions are used to be executed by the processor to implement the helium leak detection method for a switch cabinet according to any one of claims 1 to 9.

Citation Information

Patent Citations

  • Airtight container leakage positioning and measuring system based on ultrasonic detection

    CN111551321A

  • Sealing leakage rate detection method, device and equipment of connecting structure and medium

    CN115541131A

  • Method for detecting leakage of semiconductor in vacuum environment

    CN117129150A

  • Distributed industrial energy operation optimization platform automatically constructing intelligent models and algorithms

    US11487273B1

  • Intelligent analysis system using magnetic flux leakage data in pipeline inner inspection

    US20200210826A1

Cited By

  • Disease inspection method and system for buried water delivery pressure steel pipe

    CN119149964A

  • Method and system for inspecting defects of buried water pressure steel pipe

    CN119149964B

  • Graph neural network-based power grid dispatching decision-making method and large model

    CN119294872A

  • Dynamic course planning method based on adaptive learning algorithm

    CN119515625A

  • Ecological environment multi-source data intelligent acquisition and analysis system and method thereof

    CN119537810A