Power equipment fault prediction and diagnosis system based on artificial intelligence algorithm

Abstract

The invention relates to the technical field of artificial intelligence. The power equipment fault prediction and diagnosis method based on the artificial intelligence algorithm comprises the steps of performing multi-modal data acquisition and preprocessing on a vibration signal, a temperature signal, an electromagnetic field signal and a partial discharge signal of power equipment, and generating a preprocessed multi-modal feature vector; performing dynamic time reference axis construction processing on the preprocessed multi-modal feature vector to generate multi-modal data after time calibration; semantic enhancement feature fusion processing is carried out according to the multi-modal data, and global semantic consistency features are generated; performing fault prediction and diagnosis model training processing to generate a fault type classification result and a root cause tracing result; and based on a fault type classification result and a root cause tracing result, carrying out visual interaction interface generation processing, and outputting a decision view comprising a fault probability hot area distribution diagram and an economic disposal strategy so as to solve the problems of multi-modal data time sequence dislocation and physical association missing in related technologies.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence algorithm-based power equipment fault prediction and diagnosis system. Background Art

[0002] With the improvement of the intelligence level of power equipment, fault prediction and diagnosis technology of key equipment such as transformers, circuit breakers, and cables has become the core link to ensure the safe operation of the power grid.

[0003] However, due to the lack of dynamic time reference axis construction for multimodal data, related technologies lead to a misalignment in the timing of causal events and physical state evolution; at the same time, cross-modal feature fusion ignores the embedding of physical mechanisms such as heat conduction equations and electromagnetic field coupling, resulting in poor interpretability of fault prediction models and an increase in the missed detection rate of early signs. Summary of the Invention

[0004] Based on this, it is necessary to provide an electric power equipment fault prediction and diagnosis system based on artificial intelligence algorithms to address the above-mentioned technical issues, so as to solve the problems of multimodal data timing misalignment and lack of physical correlation in related technologies, so as to improve the time calibration accuracy of the fault prediction model, enhance the physical interpretability of cross-modal features and reduce the missed detection rate of early signs.

[0005] This application provides a power equipment fault prediction and diagnosis system based on artificial intelligence algorithms, which includes:

[0006] The multimodal acquisition and preprocessing module is used to perform multimodal data acquisition and preprocessing on the vibration signal, temperature signal, electromagnetic field signal and partial discharge signal of the power equipment to generate the preprocessed multimodal feature vector;

[0007] The dynamic time calibration module is used to construct a dynamic time reference axis for the pre-processed multimodal feature vector based on the device physical model and the causal event map to generate time-calibrated multimodal data;

[0008] The semantic feature fusion module is used to perform semantic enhancement feature fusion processing on the time-calibrated multimodal data according to physical association rules to generate global semantic consistency features;

[0009] The fault prediction and diagnosis module is used to train fault prediction and diagnosis models based on global semantic consistency features, generating fault type classification results and root cause tracing results;

[0010] The visualization decision generation module is used to generate and process a visualization interactive interface based on the fault type classification results and root cause tracing results, and output a decision view including a fault probability hot zone distribution map and an economic disposal strategy.

[0011] Furthermore, a visual interactive interface is generated based on the fault type classification results and root cause tracing results, and the output includes a fault probability hotspot distribution map and a decision view of the economic treatment strategy, including:

[0012] Use the following formula to perform three-dimensional spatial mapping of the equipment based on the fault type classification results and root cause tracing results to generate a fault probability hotspot distribution map:

[0013]

[0014] Among them, G i (x, y, z) represents the three-dimensional Gaussian distribution value of the i-th fault sample at the spatial point (x, y, z), X represents the spatial coordinate vector [x, y, z], μ i represents the spatial position vector of the i-th fault sample, Σ i represents the covariance matrix of the i-th fault sample, |Σ i | represents the determinant of the covariance matrix;

[0015] Based on the fault probability hot zone distribution map, the risk level is marked according to the fault propagation path, and a color gradient visualization view is generated;

[0016] Based on the color gradient visualization view, combined with the real-time electricity price data of the power market, economic optimization processing is performed to generate a decision view.

[0017] Furthermore, based on the fault probability hotspot distribution map, the risk level is labeled by fault propagation path, and a color gradient visualization is generated, including:

[0018] Based on the fault probability hot zone distribution map, the fault propagation path is dynamically tracked to generate a fault propagation path topology map;

[0019] Use the following formula to assign node risk weights based on the fault propagation path topology diagram and generate a risk level weight distribution table:

[0020]

[0021] RW i =α×R i +β×CI i +γ×S i

[0022] Among them, R i represents the risk weight of node i, n represents the total number of nodes connected to node i, ω j represents the inherent risk weight of node j, P ij represents the probability of fault propagation from node j to node i, CI irepresents the centrality index of node i, m represents the total number of nodes in the topological graph, and d ij represents the shortest path distance from node i to node j, f j represents the failure frequency of node j, RW i represents the comprehensive risk weight of node i, α represents the weight coefficient of risk weight, β represents the weight coefficient of centrality index, γ represents the weight coefficient of node importance, and α+β+γ=1, S i represents the importance score of node i;

[0023] Based on the risk level weight distribution table, gradient mapping is performed using color coding rules to generate a color gradient visualization view.

[0024] Furthermore, based on the color gradient visualization view, combined with real-time electricity price data from the power market, economic optimization processing is performed to generate a decision view, including:

[0025] Based on the color gradient visualization view, risk nodes are prioritized and a list of key risk nodes is generated;

[0026] Based on the list of key risk nodes, combined with the real-time electricity price data of the power market, the electricity price correlation processing of the time period is carried out to generate the cost impact assessment matrix;

[0027] Based on the cost impact assessment matrix, a multi-objective optimization algorithm is used to perform strategic trade-offs and generate a decision view.

[0028] Furthermore, based on the fault type classification results and root cause tracing results, three-dimensional equipment space mapping is performed to generate a fault probability hotspot distribution map, including:

[0029] Based on the fault type classification results and root cause tracing results, spatial coordinate mapping is performed through the 3D physical model of the equipment to generate a 3D fault coordinate set;

[0030] Based on the three-dimensional fault coordinate set, a dynamic weight distribution matrix is ​​generated by dynamically allocating the fault type weights.

[0031] Based on the dynamic weight distribution matrix, the probability density calculation is performed through the hot zone aggregation algorithm to generate a fault probability hot zone distribution map.

[0032] Furthermore, semantic enhancement feature fusion processing is performed on the time-calibrated multimodal data according to physical association rules to generate global semantic consistency features, including:

[0033] Based on physical association rules, a multimodal physical knowledge graph is generated through the construction and processing of the device physical knowledge graph;

[0034] Based on the time-calibrated multimodal data and the multimodal physical knowledge graph, graph structure conversion is performed to generate graph data including node features and physical association edges;

[0035] Based on graph data, features are dynamically aggregated through a cross-modal attention mechanism to generate global semantic consistency features.

[0036] Furthermore, based on the time-calibrated multimodal data and the multimodal physical knowledge graph, graph structure conversion processing is performed to generate graph data including node features and physical association edges, including:

[0037] Based on the time-calibrated multimodal data, node feature vectors are generated through physical attribute encoding processing;

[0038] Based on the multimodal physical knowledge graph, the physical association edge weight matrix is ​​generated through dynamic relationship modeling.

[0039] Based on the node feature vector and the physical association edge weight matrix, the graph topology structure is generated and processed to generate graph data containing node features and physical association edges.

[0040] Furthermore, based on the multimodal physical knowledge graph, a physical association edge weight matrix is ​​generated through dynamic relationship modeling, including:

[0041] Based on the multimodal physical knowledge graph, the physical relationship subgraph associated with the equipment operating status is generated through dynamic analysis of physical relationships;

[0042] Based on the physical relationship subgraph, real-time device status matching is performed through dynamic weight calculation rules to generate a dynamic association weight parameter set;

[0043] Based on the dynamic association weight parameter set, the weight matrix is ​​reconstructed to generate the physical association edge weight matrix.

[0044] Furthermore, a dynamic time reference axis is constructed on the pre-processed multimodal feature vector based on the device physical model and the causal event graph to generate time-calibrated multimodal data, including:

[0045] Based on the physical model of the equipment, the theoretical time axis is generated through the physical state evolution equation to generate the theoretical state evolution time reference axis;

[0046] Based on the causal event graph, the pre-processed multimodal feature vector is processed to identify causal event nodes and generate an event-driven temporal dependency graph;

[0047] Based on the theoretical state evolution time reference axis and time series dependency graph, multimodal data calibration is performed through dynamic timestamp remapping rules to generate time-calibrated multimodal data.

[0048] Furthermore, the fault prediction and diagnosis model training process is performed based on the global semantic consistency features to generate fault type classification results and root cause tracing results, including:

[0049] Based on the global semantic consistency features, a hybrid training model is constructed to generate supervised learning branches and unsupervised learning branches.

[0050] The supervised learning branch is used to classify fault types and generate classification results including overtemperature and insulation breakdown.

[0051] Perform abnormal pattern detection processing through the unsupervised learning branch to generate abnormal features including deviations from the normal state baseline;

[0052] Based on the classification results and abnormal characteristics, the fault propagation path is inverted through the graph structure root cause tracing network to generate fault type classification results and root cause tracing results.

[0053] The technical solution provided by the present application includes the following technical effects: by providing a power equipment fault prediction and diagnosis system based on artificial intelligence algorithm, the system includes: a multimodal acquisition and preprocessing module for performing multimodal data acquisition and preprocessing on the vibration signal, temperature signal, electromagnetic field signal and partial discharge signal of the power equipment to generate a preprocessed multimodal feature vector; a dynamic time calibration module for performing dynamic time reference axis construction processing on the preprocessed multimodal feature vector based on the equipment physical model and causal event map to generate time-calibrated multimodal data; a semantic feature fusion module for fusion processing the time-calibrated multimodal data according to physical association rules. The system performs semantic enhancement feature fusion processing based on the global semantic consistency feature to generate global semantic consistency feature; the fault prediction and diagnosis module is used to train the fault prediction and diagnosis model based on the global semantic consistency feature to generate fault type classification results and root cause tracing results; the visualization decision generation module is used to generate and process the visualization interactive interface based on the fault type classification results and root cause tracing results, and output a decision view including the fault probability hot zone distribution map and the economic disposal strategy, so as to solve the problems of multimodal data time series misalignment and lack of physical association in related technologies, so as to improve the time calibration accuracy of the fault prediction model, enhance the physical interpretability of cross-modal features and reduce the early sign missed detection rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0055] Figure 1 This is a structural diagram of an electric power equipment fault prediction and diagnosis system based on an artificial intelligence algorithm in one embodiment of the present invention;

[0056] Figure 2 A flowchart of an embodiment of the present invention, which is based on a color gradient visualization view and combines real-time electricity price data from the power market to perform economic optimization processing and generate a decision view;

[0057] Figure 3 The present invention provides a flowchart for constructing a dynamic time reference axis for preprocessed multimodal feature vectors based on a device physical model and a causal event graph to generate time-calibrated multimodal data in one embodiment of the present invention. DETAILED DESCRIPTION

[0058] In order to make the above-mentioned purposes, features and advantages of the present application more clearly understood, the specific implementation methods of the present application are described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to fully understand the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar improvements without violating the connotation of the application. Therefore, the present application is not limited to the specific embodiments disclosed below.

[0059] like Figure 1 As shown, the present application provides a power equipment fault prediction and diagnosis system 100 based on artificial intelligence algorithm, which includes:

[0060] The multimodal acquisition and preprocessing module 101 is used to perform multimodal data acquisition and preprocessing on vibration signals, temperature signals, electromagnetic field signals and partial discharge signals of the power equipment to generate preprocessed multimodal feature vectors.

[0061] Specifically, data is collected from vibration, temperature, electromagnetic field, and partial discharge signals of power equipment. A sensor network is used to obtain raw data from these different signal sources, ensuring data integrity and accuracy. This data is then initially processed, including filtering, denoising, and normalization, to improve its quality and usability.

[0062] Afterwards, feature extraction is performed on different types of signals. For vibration signals, frequency domain features are extracted; for temperature signals, time-varying features are extracted; for electromagnetic field signals, amplitude and phase features are extracted; and for partial discharge signals, discharge intensity and frequency features are extracted. The extracted signal features are fused and combined into multimodal feature vectors according to specific rules and formats, providing basic data support for subsequent fault prediction and diagnosis.

[0063] The dynamic time calibration module 102 is used to construct a dynamic time reference axis for the pre-processed multimodal feature vector based on the device physical model and the causal event map to generate time-calibrated multimodal data.

[0064] Specifically, based on the physical model of the device, the time reference axis of its theoretical state evolution is determined; then, the causal event graph is used to identify the causal event nodes in the multimodal feature vector to generate a timing dependency graph; then, through the dynamic timestamp remapping rule, the preprocessed multimodal feature vector is mapped to the theoretical state evolution time reference axis to achieve time calibration of the multimodal data.

[0065] The semantic feature fusion module 103 is used to perform semantic enhancement feature fusion processing on the time-calibrated multimodal data according to physical association rules to generate global semantic consistency features.

[0066] Specifically, based on the physical characteristics and operating mechanisms of the equipment, a corresponding physical knowledge graph is constructed to clarify the association rules between various physical quantities. The time-calibrated multimodal data is then mapped onto this physical knowledge graph. Through graph structure conversion, the data from different modalities is transformed into a graph data format consisting of node features and physical association edges. The node features represent the key characteristics of the data from each modality, while the physical association edges reflect the inherent physical connections between them.

[0067] Afterwards, the cross-modal attention mechanism is used to dynamically aggregate features of the graph data. During this process, features of different modalities are assigned different weights and fused according to their importance and relevance in physical associations, thereby generating features that can comprehensively reflect the device status and have global semantic consistency, namely global semantic consistency features, providing a more physically interpretable and comprehensive feature basis for subsequent fault prediction and diagnosis.

[0068] The fault prediction and diagnosis module 104 is used to perform fault prediction and diagnosis model training based on global semantic consistency features, and generate fault type classification results and root cause tracing results.

[0069] Specifically, a hybrid training model is constructed, consisting of supervised and unsupervised learning branches. In the supervised learning branch, training is performed using labeled global semantically consistent feature samples. By learning the association between features and fault types, a fault type classification model is generated, enabling more accurate classification of fault types such as overtemperature and insulation breakdown. In the unsupervised learning branch, training is performed on unlabeled global semantically consistent feature samples. Through cluster analysis and pattern recognition, abnormal patterns in the equipment's operating status are detected, identifying features that deviate from the normal baseline.

[0070] Afterwards, the outputs of supervised learning and unsupervised learning are combined, and a graph-structured root cause tracing network is used to invert the fault propagation path. By analyzing the correlation between fault types and abnormal patterns, fault type classification results and root cause tracing results are generated, thereby achieving a more comprehensive prediction and diagnosis of faults.

[0071] The visualization decision generation module 105 is used to generate a visualization interactive interface based on the fault type classification results and the root cause tracing results, and output a decision view including a fault probability hot zone distribution map and an economic disposal strategy.

[0072] Specifically, based on the fault type classification and root cause tracing results, spatial coordinate mapping is performed using the equipment's 3D physical model to generate a 3D fault coordinate set containing fault location information. This 3D coordinate set is then weighted according to the fault type and root cause tracing information to generate a dynamic weight distribution matrix. This matrix reflects the probability and importance of faults at different locations. A hotspot aggregation algorithm is then used to calculate the probability density of the dynamic weight distribution matrix, generating a fault probability hotspot distribution map that visually displays the degree of fault risk in each area of ​​the equipment.

[0073] At the same time, the system combines real-time electricity market price data and fault information to conduct economic evaluations, analyze the costs and benefits of different disposal strategies, and generate a decision-making view for economic disposal strategies. The fault probability hotspot distribution map and the economic disposal strategy decision view are integrated into a visual interactive interface. Through the generated processing of the visual interactive interface, the output includes the fault probability hotspot distribution map and the economic disposal strategy decision view, which users can use to make fault disposal decisions.

[0074] An embodiment of the present application provides a power equipment fault prediction and diagnosis system based on an artificial intelligence algorithm, which includes: a multimodal acquisition and preprocessing module for performing multimodal data acquisition and preprocessing on vibration signals, temperature signals, electromagnetic field signals, and partial discharge signals of power equipment to generate preprocessed multimodal feature vectors; a dynamic time calibration module for performing dynamic time reference axis construction processing on the preprocessed multimodal feature vectors based on the equipment physical model and causal event map to generate time-calibrated multimodal data; a semantic feature fusion module for performing semantic enhancement on the time-calibrated multimodal data according to physical association rules Feature fusion processing generates global semantic consistency features; the fault prediction and diagnosis module is used to train fault prediction and diagnosis models based on global semantic consistency features, and generate fault type classification results and root cause tracing results; the visualization decision generation module is used to generate and process visual interactive interfaces based on fault type classification results and root cause tracing results, and output decision views including fault probability hot zone distribution maps and economic disposal strategies, in order to solve the problems of multimodal data time series misalignment and lack of physical association in related technologies, so as to improve the time calibration accuracy of fault prediction models, enhance the physical interpretability of cross-modal features and reduce the missed detection rate of early signs.

[0075] Furthermore, a visual interactive interface is generated based on the fault type classification results and root cause tracing results, and the output includes a fault probability hotspot distribution map and a decision view of the economic treatment strategy, including:

[0076] Use the following formula to perform three-dimensional spatial mapping of the equipment based on the fault type classification results and root cause tracing results to generate a fault probability hotspot distribution map:

[0077]

[0078] Among them, G i (x, y, z) represents the three-dimensional Gaussian distribution value of the i-th fault sample at the spatial point (x, y, z), X represents the spatial coordinate vector [x, y, z], μ i represents the spatial position vector of the i-th fault sample, Σ i represents the covariance matrix of the i-th fault sample, |Σ i | represents the determinant of the covariance matrix;

[0079] Based on the fault probability hot zone distribution map, the risk level is marked according to the fault propagation path, and a color gradient visualization view is generated;

[0080] Based on the color gradient visualization view, combined with the real-time electricity price data of the power market, economic optimization processing is performed to generate a decision view.

[0081] Specifically, based on the fault type classification and root cause tracing results, the 3D physical model of the equipment is used for spatial coordinate mapping, generating a 3D fault coordinate set that includes fault location information. Dynamic weights are then assigned to the 3D fault coordinate set based on the fault type and root cause tracing information, generating a dynamic weight distribution matrix that reflects the probability and importance of faults at different locations. A hotspot aggregation algorithm is then used to calculate the probability density of the dynamic weight distribution matrix, generating a fault probability hotspot distribution map that visually displays the degree of fault risk in each area of ​​the equipment.

[0082] After generating a fault probability hotspot distribution map, the system analyzes fault propagation paths and labels the hotspot distribution map with risk levels. Using color gradient technology, the different risk levels are visually displayed through color variations, generating a color gradient visualization. This visualization is then combined with real-time electricity price data from the power market to perform economic optimization on the color gradient visualization. The system then evaluates the costs and benefits of different response strategies, generating a decision view that includes the fault probability hotspot distribution map and economic response strategies. This view is presented to users through a visual interactive interface to assist them in making fault response decisions.

[0083] Furthermore, based on the fault probability hotspot distribution map, the risk level is labeled by fault propagation path, and a color gradient visualization is generated, including:

[0084] Based on the fault probability hot zone distribution map, the fault propagation path is dynamically tracked to generate a fault propagation path topology map;

[0085] Use the following formula to assign node risk weights based on the fault propagation path topology diagram and generate a risk level weight distribution table:

[0086]

[0087] RW i =α×R i +β×CI i +γ×S i

[0088] Among them, R i represents the risk weight of node i, n represents the total number of nodes connected to node i, ω j represents the inherent risk weight of node j, P ij represents the probability of fault propagation from node j to node i, CI i represents the centrality index of node i, m represents the total number of nodes in the topological graph, and d ij represents the shortest path distance from node i to node j, f j represents the failure frequency of node j, RW irepresents the comprehensive risk weight of node i, α represents the weight coefficient of risk weight, β represents the weight coefficient of centrality index, γ represents the weight coefficient of node importance, and α+β+γ=1, S i represents the importance score of node i;

[0089] Based on the risk level weight distribution table, gradient mapping is performed using color coding rules to generate a color gradient visualization view.

[0090] Specifically, using the fault probability hotspot distribution map as a foundation and combining it with the device's physical architecture and operational characteristics, the fault propagation path is dynamically tracked. This step aims to identify and depict the possible propagation paths of the fault within the device or system, thereby generating a fault propagation path topology map. This topology map not only shows the possible fault propagation paths but also illustrates the connection relationships between nodes and the propagation direction.

[0091] Next, node risk weights are assigned. This process considers multiple risk factors, such as the inherent risk attributes of the node, the probability of fault propagation between nodes, the node's centrality within the network, the shortest path distance between nodes, and the frequency of node faults. Through a systematic analysis of these factors, a risk weight is calculated for each node, which in turn generates a risk level weight distribution table. This distribution table assigns each node a risk weight value that reflects the node's potential risk in fault propagation.

[0092] Next, based on the risk level weight distribution table, color coding rules are applied to perform gradient mapping. This involves assigning colors to different risk levels based on their weights, generating a color gradient. This visualization process generates a color gradient visualization that presents fault risk levels through intuitive color changes, allowing users to quickly identify high-risk areas and critical fault propagation paths. This visualization not only improves the intuitiveness of fault diagnosis but also provides a valuable visual reference for subsequent maintenance decisions.

[0093] like Figure 2 As shown in the figure, based on the color gradient visualization view, combined with the real-time electricity price data of the power market, economic optimization processing is performed to generate a decision view, including:

[0094] S201: Based on the color gradient visualization view, risk nodes are prioritized and a list of key risk nodes is generated;

[0095] S202: Based on the key risk node list, combined with the real-time electricity price data of the power market, the electricity price correlation processing of the time period is performed to generate a cost impact assessment matrix;

[0096] S203: Based on the cost impact assessment matrix, a multi-objective optimization algorithm is used to perform strategy trade-off processing and generate a decision view.

[0097] Specifically, in step S201, the color gradient visualization view is analyzed to identify the risk levels represented by different color gradients, and each risk node is prioritized according to the above risk level, thereby generating a list of key risk nodes to clarify the key objects for subsequent economic analysis.

[0098] Next, in step S202, the resulting list of key risk nodes is combined with real-time electricity price data from the power market to perform time-period price correlation processing. This process involves considering electricity price fluctuations over different time periods and analyzing the impact of price changes on the maintenance or repair costs of each key risk node. This process then generates a cost impact assessment matrix that shows the cost changes associated with handling failures at different risk nodes over different time periods.

[0099] Next, in step S203, a multi-objective optimization algorithm is used to weigh various possible fault handling strategies based on the cost impact assessment matrix. This algorithm comprehensively considers multiple objectives, such as minimizing costs and maximizing fault handling effectiveness, to find a balance among the numerous strategies. This generates a decision view that clearly presents the pros and cons of each strategy under different constraints, providing decision makers with a scientific and rational basis for developing cost-effective and reliable power equipment fault handling solutions.

[0100] Furthermore, based on the fault type classification results and root cause tracing results, three-dimensional equipment space mapping is performed to generate a fault probability hotspot distribution map, including:

[0101] Based on the fault type classification results and root cause tracing results, spatial coordinate mapping is performed through the 3D physical model of the equipment to generate a 3D fault coordinate set;

[0102] Based on the three-dimensional fault coordinate set, a dynamic weight distribution matrix is ​​generated by dynamically allocating the fault type weights.

[0103] Based on the dynamic weight distribution matrix, the probability density calculation is performed through the hot zone aggregation algorithm to generate a fault probability hot zone distribution map.

[0104] Specifically, the fault type classification results are combined with the root cause tracing results, and the fault information is mapped to spatial coordinates using the device's 3D physical model. This step maps the abstract fault data to the device's actual physical location, generating a 3D fault coordinate set containing the fault location information, laying the foundation for subsequent spatial analysis.

[0105] Next, a dynamic weight is assigned to each coordinate point in the three-dimensional fault coordinate set based on detailed information about the fault type and root cause. This process considers factors such as the severity and frequency of the fault type and the impact of the root cause. This weighting highlights high-risk areas and generates a dynamic weight distribution matrix that quantitatively describes the degree of fault risk at different locations on the equipment.

[0106] The dynamic weight distribution matrix is ​​then processed using a heatmap aggregation algorithm. This algorithm analyzes the weight distribution and calculates the failure probability density for each area of ​​the equipment. This algorithm integrates the scattered failure data into intuitive visual information, generating a failure probability heatmap. This map uses color shading or light / dark variations to visually display the failure probability of different parts of the equipment, helping maintenance personnel more quickly locate high-risk areas and thus achieving visual management of power equipment failure risks.

[0107] Furthermore, semantic enhancement feature fusion processing is performed on the time-calibrated multimodal data according to physical association rules to generate global semantic consistency features, including:

[0108] Based on physical association rules, a multimodal physical knowledge graph is generated through the construction and processing of the device physical knowledge graph;

[0109] Based on the time-calibrated multimodal data and the multimodal physical knowledge graph, graph structure conversion is performed to generate graph data including node features and physical association edges;

[0110] Based on graph data, features are dynamically aggregated through a cross-modal attention mechanism to generate global semantic consistency features.

[0111] Specifically, a physical knowledge graph of the device is constructed based on physical association rules. This step integrates the physical characteristics and operating mechanisms of the device to generate a multimodal physical knowledge graph, providing a structured knowledge framework for subsequent data fusion.

[0112] Afterward, the time-calibrated multimodal data is mapped onto a multimodal physical knowledge graph for graph structure conversion. This step transforms the data from different modalities into a graph, where node features represent the key characteristics of each modal data, and physical edges reflect the inherent physical connections between them.

[0113] The cross-modal attention mechanism then dynamically aggregates features from the graph data. Based on the importance and relevance of each modal feature in the physical context, the cross-modal attention mechanism dynamically adjusts feature weights and fuses them together to generate a globally semantically consistent feature. This feature comprehensively reflects the device state, offering enhanced physical interpretability and consistency, providing a foundation for subsequent fault prediction and diagnosis.

[0114] Furthermore, based on the time-calibrated multimodal data and the multimodal physical knowledge graph, graph structure conversion processing is performed to generate graph data including node features and physical association edges, including:

[0115] Based on the time-calibrated multimodal data, node feature vectors are generated through physical attribute encoding processing;

[0116] Based on the multimodal physical knowledge graph, the physical association edge weight matrix is ​​generated through dynamic relationship modeling.

[0117] Based on the node feature vector and the physical association edge weight matrix, the graph topology structure is generated and processed to generate graph data containing node features and physical association edges.

[0118] Specifically, physical attribute encoding is performed on the time-calibrated multimodal data. This process involves encoding the physical attributes of different modal data, mapping physical quantities such as vibration, temperature, and electromagnetic fields into feature vectors. This generates node feature vectors that characterize the key characteristics of each modal data, laying the foundation for subsequent graph data construction.

[0119] Next, dynamic relationship modeling is performed based on the multimodal physical knowledge graph. This step focuses on constructing a physical association edge weight matrix. By dynamically analyzing the physical entities and relationships in the knowledge graph, the strength of the associations between different physical quantities is quantified, generating a physical association edge weight matrix that intuitively reflects the correlation and influence between the various physical quantities.

[0120] The graph topology is then generated by combining the node feature vectors with the physical edge weight matrix. By integrating the node feature vectors with the physical edge weight matrix, a complete graph data structure is constructed, where nodes represent the key features of multimodal data and edges represent the relationships between physical quantities and their weights. This generates graph data containing both node features and physical edges, providing structured input for subsequent graph neural network analysis or other graph-based processing.

[0121] Furthermore, based on the multimodal physical knowledge graph, a physical association edge weight matrix is ​​generated through dynamic relationship modeling, including:

[0122] Based on the multimodal physical knowledge graph, the physical relationship subgraph associated with the equipment operating status is generated through dynamic analysis of physical relationships;

[0123] Based on the physical relationship subgraph, real-time device status matching is performed through dynamic weight calculation rules to generate a dynamic association weight parameter set;

[0124] Based on the dynamic association weight parameter set, the weight matrix is ​​reconstructed to generate the physical association edge weight matrix.

[0125] Specifically, through dynamic parsing of physical relationships, the physical relationships within the multimodal physical knowledge graph are analyzed and extracted in real time. This process aims to identify and construct physical relationship subgraphs that are closely related to the current operating state of the equipment. These subgraphs contain correlations between different physical quantities, such as the interactions between vibration and temperature, or electromagnetic fields and partial discharges.

[0126] Next, dynamic weight calculation rules are used to perform real-time device state matching on the physical relationships within the physical relationship subgraph. This step dynamically calculates the weight parameters for each physical relationship by analyzing the changes in physical quantities under the current device operating state, generating a set of dynamic association weight parameters. The dynamic weight calculation rules take into account the changing trends of physical quantities, historical data, and contextual information about the device operating state to quantify the importance of each physical relationship.

[0127] Next, based on the dynamic association weight parameter set, a weight matrix reconstruction process is performed to generate a physical association edge weight matrix. This step integrates the weight parameters in the dynamic association weight parameter set into a matrix structure to generate the physical association edge weight matrix. Each element in this matrix represents the strength of the association between different physical quantities, providing a foundation for subsequent graph data processing and analysis. It also provides a quantitative basis for physical associations for cross-modal feature fusion.

[0128] like Figure 3 As shown, based on the device physical model and causal event graph, the pre-processed multimodal feature vector is subjected to dynamic time reference axis construction to generate time-calibrated multimodal data, including:

[0129] S301: Based on the physical model of the device, a theoretical time axis is generated by using the physical state evolution equation to generate a theoretical state evolution time reference axis;

[0130] S302: Based on the causal event graph, perform causal event node identification processing on the pre-processed multimodal feature vector to generate an event-driven temporal dependency graph;

[0131] S303: Based on the theoretical state evolution time reference axis and the time sequence dependency graph, multimodal data calibration processing is performed through dynamic timestamp remapping rules to generate time-calibrated multimodal data.

[0132] Specifically, in step S301, based on the physical model of the device, the physical state evolution equation is used to simulate and predict the theoretical operating state of the device to generate a theoretical state evolution time reference axis, which provides a reference for the physical state of the device changing over time under ideal conditions.

[0133] Afterwards, in step S302, based on the causal event graph, the pre-processed multimodal feature vector is deeply analyzed to identify the causal event nodes contained therein, and an event-driven temporal dependency graph is constructed. This graph depicts the sequence and causal relationship between different events, reflecting the time series relationship of key events during the operation of the equipment.

[0134] Then, in step S303, the multimodal data is calibrated using dynamic timestamp remapping rules, combining the theoretical state evolution time reference axis and the timing dependency graph. By comparing the time difference between the actual data and the theoretical reference, the timestamps of the multimodal feature vectors are adjusted to more accurately align the data of different modalities in the time dimension, ensuring that each modal data more accurately reflects the actual state of the device at the same point in time. This generates time-calibrated multimodal data, providing temporally consistent data support for subsequent fault prediction and diagnosis.

[0135] Furthermore, the fault prediction and diagnosis model training process is performed based on the global semantic consistency features to generate fault type classification results and root cause tracing results, including:

[0136] Based on the global semantic consistency features, a hybrid training model is constructed to generate supervised learning branches and unsupervised learning branches.

[0137] The supervised learning branch is used to classify fault types and generate classification results including overtemperature and insulation breakdown.

[0138] Perform abnormal pattern detection processing through the unsupervised learning branch to generate abnormal features including deviations from the normal state baseline;

[0139] Based on the classification results and abnormal characteristics, the fault propagation path is inverted through the graph structure root cause tracing network to generate fault type classification results and root cause tracing results.

[0140] Specifically, a hybrid training model is constructed using global semantic consistency features. This step involves building two branches: a supervised learning branch and an unsupervised learning branch. The supervised learning branch focuses on learning from labeled data, capable of identifying and classifying specific fault types such as overtemperature and insulation breakdown. The unsupervised learning branch processes unlabeled data, aiming to discover hidden patterns and abnormal characteristics in the data and identify abnormalities that deviate from normal operation.

[0141] The supervised learning branch then classifies the fault type. This branch uses labeled data of known fault types for training, learns the characteristic patterns of different fault types, and generates corresponding classification results, clearly indicating the specific fault type that the device may encounter.

[0142] Meanwhile, the unsupervised learning branch performs abnormal pattern detection. Without predefined fault labels, this branch analyzes the inherent structure and distribution of the data to identify abnormal features that differ significantly from normal operating conditions, thereby discovering potential unknown fault modes.

[0143] Then, combining the classification results from the supervised learning branch with the anomaly features from the unsupervised learning branch, a graph-structured root cause tracing network is used to inverse the fault propagation path. This network analyzes the causal relationship and propagation path between fault type and anomaly features, tracing the root cause of the fault. This generates comprehensive fault type classification and root cause tracing results, providing more precise guidance for equipment maintenance and troubleshooting.

[0144] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0145] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0146] The above-described embodiments merely represent several implementation methods of the embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the concept of the embodiments of the present application, and these modifications and improvements fall within the scope of protection of the embodiments of the present application.

Claims

1. Power equipment fault prediction and diagnosis system based on artificial intelligence algorithm, characterized by: The system comprises: The multimodal acquisition and preprocessing module is used to perform multimodal data acquisition and preprocessing on the vibration signal, temperature signal, electromagnetic field signal and partial discharge signal of the power equipment to generate the preprocessed multimodal feature vector; A dynamic time calibration module is used to construct a dynamic time reference axis for the pre-processed multimodal feature vector based on the device physical model and the causal event map to generate time-calibrated multimodal data; A semantic feature fusion module, configured to perform semantic enhancement feature fusion processing on the time-calibrated multimodal data according to physical association rules to generate global semantic consistency features; A fault prediction and diagnosis module is used to perform fault prediction and diagnosis model training based on the global semantic consistency features, and generate fault type classification results and root cause tracing results; The visualization decision generation module is used to generate a visualization interactive interface based on the fault type classification results and root cause tracing results, and output a decision view including a fault probability hot zone distribution map and an economic disposal strategy.

2. The power equipment fault prediction and diagnosis system based on artificial intelligence algorithm according to claim 1 is characterized in that: The generation of a visual interactive interface based on the fault type classification result and the root cause tracing result, outputting a decision view including a fault probability hot zone distribution map and an economic disposal strategy, includes: Use the following formula to perform three-dimensional spatial mapping of the equipment based on the fault type classification results and root cause tracing results to generate a fault probability hotspot distribution map: Among them, G i (x, y, z) represents the three-dimensional Gaussian distribution value of the i-th fault sample at the spatial point (x, y, z), X represents the spatial coordinate vector [x, y, z], μ i represents the spatial position vector of the i-th fault sample, Σ i represents the covariance matrix of the i-th fault sample, |Σ i | represents the determinant of the covariance matrix; Based on the fault probability hot zone distribution map, risk level labeling is performed through the fault propagation path to generate a color gradient visualization view; Based on the color gradient visualization view, economic optimization processing is performed in combination with real-time electricity price data of the power market to generate the decision view.

3. The power equipment fault prediction and diagnosis system based on artificial intelligence algorithm according to claim 2 is characterized in that: The method of performing risk level labeling processing based on the fault probability hot zone distribution map and the fault propagation path to generate a color gradient visualization view includes: Based on the fault probability hot zone distribution map, a fault propagation path topology map is generated by dynamically tracking the fault propagation path; The following formula is used to perform node risk weight allocation based on the fault propagation path topology diagram to generate a risk level weight distribution table: RW i =α×R i +β×CI i +γ×S i Among them, R i represents the risk weight of node i, n represents the total number of nodes connected to node i, ω j represents the inherent risk weight of node j, P ij represents the probability of fault propagation from node j to node i, CI i represents the centrality index of node i, m represents the total number of nodes in the topological graph, and d ij represents the shortest path distance from node i to node j, f j represents the failure frequency of node j, RW i represents the comprehensive risk weight of node i, α represents the weight coefficient of risk weight, β represents the weight coefficient of centrality index, γ represents the weight coefficient of node importance, and α+β+γ=1, S i represents the importance score of node i; Based on the risk level weight distribution table, gradient mapping processing is performed using color coding rules to generate the color gradient visualization view.

4. The power equipment fault prediction and diagnosis system based on artificial intelligence algorithm according to claim 2 is characterized in that: The economic optimization process based on the color gradient visualization view and combined with the real-time electricity price data of the power market to generate the decision view includes: Based on the color gradient visualization view, risk nodes are prioritized to generate a list of key risk nodes; Based on the key risk node list, combined with the real-time electricity price data of the power market, time period electricity price correlation processing is performed to generate a cost impact assessment matrix; Based on the cost impact assessment matrix, a strategy trade-off process is performed through a multi-objective optimization algorithm to generate the decision view.

5. The power equipment fault prediction and diagnosis system based on artificial intelligence algorithm according to claim 2 is characterized in that: The three-dimensional spatial mapping of the equipment is performed based on the fault type classification result and the root cause tracing result to generate a fault probability hot zone distribution map, including: Based on the fault type classification results and root cause tracing results, spatial coordinate mapping processing is performed through the three-dimensional physical model of the equipment to generate a three-dimensional fault coordinate set; Based on the three-dimensional fault coordinate set, a dynamic weight distribution matrix is ​​generated by performing a dynamic weight distribution process on the fault type; Based on the dynamic weight distribution matrix, probability density calculation processing is performed through a hot zone aggregation algorithm to generate the fault probability hot zone distribution map.

6. The power equipment fault prediction and diagnosis system based on artificial intelligence algorithm according to claim 1 is characterized in that: The step of performing semantic enhancement feature fusion processing on the time-calibrated multimodal data according to physical association rules to generate global semantic consistency features includes: Based on the physical association rules, a multimodal physical knowledge graph is generated through device physical knowledge graph construction and processing; Performing graph structure conversion processing based on the time-calibrated multimodal data and the multimodal physical knowledge graph to generate graph data including node features and physical association edges; Based on the graph data, dynamic feature aggregation processing is performed through a cross-modal attention mechanism to generate the global semantic consistency feature.

7. The power equipment fault prediction and diagnosis system based on artificial intelligence algorithm according to claim 6 is characterized in that: The graph structure conversion process is performed based on the time-calibrated multimodal data and the multimodal physical knowledge graph to generate graph data including node features and physical associated edges, including: Based on the time-calibrated multimodal data, generating a node feature vector through physical attribute encoding processing; Based on the multimodal physical knowledge graph, a physical association edge weight matrix is ​​generated through dynamic relationship modeling processing; Based on the node feature vector and the physical association edge weight matrix, a graph topology structure generation process is performed to generate the graph data including the node features and the physical association edges.

8. The power equipment fault prediction and diagnosis system based on artificial intelligence algorithm according to claim 7 is characterized in that: The method of generating a physical association edge weight matrix based on the multimodal physical knowledge graph through dynamic relationship modeling processing includes: Based on the multimodal physical knowledge graph, a physical relationship subgraph associated with the device operating status is generated through dynamic physical relationship analysis; Based on the physical relationship subgraph, real-time device state matching processing is performed through dynamic weight calculation rules to generate a dynamic association weight parameter set; Based on the dynamic association weight parameter set, a weight matrix reconstruction process is performed to generate the physical association edge weight matrix.

9. The power equipment fault prediction and diagnosis system based on artificial intelligence algorithm according to claim 1 is characterized in that: The process of constructing a dynamic time reference axis on the pre-processed multimodal feature vector based on the device physical model and the causal event graph to generate time-calibrated multimodal data includes: Based on the physical model of the device, a theoretical time axis generation process is performed through a physical state evolution equation to generate a theoretical state evolution time reference axis; Based on the causal event graph, performing causal event node identification processing on the pre-processed multimodal feature vector to generate an event-driven temporal dependency graph; Based on the theoretical state evolution time reference axis and the timing dependency graph, multimodal data calibration processing is performed through dynamic timestamp remapping rules to generate the time-calibrated multimodal data.

10. The power equipment fault prediction and diagnosis system based on artificial intelligence algorithm according to claim 1 is characterized in that: The fault prediction and diagnosis model training process based on the global semantic consistency feature to generate fault type classification results and root cause tracing results includes: Based on the global semantic consistency feature, a hybrid training model is constructed to perform supervised learning branch and unsupervised learning branch generation processing to generate supervised learning branches and unsupervised learning branches; Perform fault type classification processing through the supervised learning branch to generate classification results including overtemperature and insulation breakdown; Performing abnormal pattern detection processing through the unsupervised learning branch to generate abnormal features including deviations from a normal baseline; Based on the classification result and the abnormal characteristics, the fault propagation path inversion processing is performed through a graph structure root cause tracing network to generate the fault type classification result and the root cause tracing result.

Images