A power grid fault analysis method and system

Abstract

The present application relates to the technical field of fault diagnosis, in particular to a power grid fault analysis method and system, which comprises constructing a fusion tensor and performing dimension reduction processing on the fusion tensor to generate a fusion feature matrix; constructing a convolutional autoencoder model to detect abnormal fusion feature matrix; classifying power grid faults according to abnormal data of the abnormal feature matrix, and calculating power grid equipment fault area based on the result of power grid fault classification. The present application has the beneficial effect of ensuring the integrity and accuracy of data by using a high-precision low-rank tensor completion algorithm to complete power grid operation data. Key features are extracted from multi-dimensional data collected by multiple source sensors and a fusion feature matrix is constructed using Tucker decomposition and matrix compression technology, effectively improving the efficiency of data processing and the ability of feature expression. The application of the convolutional autoencoder model enhances the detection accuracy of abnormal features, and the improved weighted summation and hierarchical clustering-based method improves the accuracy of fault classification and positioning.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis technology, and in particular to a power grid fault analysis method and system. Background Technology

[0002] With the development of smart and digital power grids, power grid fault analysis technology based on multi-dimensional data has gradually gained attention. Traditional power grid monitoring methods mainly rely on single sensors to obtain limited operational data, which is difficult to fully reflect the operating status of the power grid. With the advancement of multi-source sensor technology, the types and dimensions of power grid operation data have increased significantly, covering physical quantities such as voltage, current, temperature, and frequency, as well as network characteristics such as network latency, bandwidth utilization, and packet loss rate. By integrating multi-dimensional data, researchers can use data-driven methods to improve the detection and diagnosis accuracy of power grid faults. In recent years, fault diagnosis models based on machine learning have been proposed, which utilize a large amount of historical data for training and can achieve rapid identification and classification of abnormal states. These technologies have played an important role in improving the safety and stability of the power grid.

[0003] Existing technologies still have many shortcomings when processing multi-dimensional data. For example, existing tensor completion algorithms have limited effectiveness in addressing the sparsity problem of multi-dimensional data and often fail to effectively fill in missing data, resulting in incomplete and unreliable analysis results. In addition, the analysis of power grid fault characteristics often lacks systematicity, leading to insufficient classification and location accuracy. Summary of the Invention

[0004] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0005] In view of the aforementioned existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides a power grid fault analysis method and system that can solve the problems mentioned in the background art.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0008] In a first aspect, the present invention provides a power grid fault analysis method, including constructing a fusion tensor and performing dimensionality reduction processing on the fusion tensor to generate a fusion feature matrix;

[0009] Construct a convolutional autoencoder model to detect anomalies and fuse feature matrices;

[0010] Power grid faults are classified based on the abnormal data in the abnormal feature matrix. Based on the results of the power grid fault classification, the fault areas of power grid equipment are calculated, and fault reports are generated.

[0011] As a preferred embodiment of the power grid fault analysis method of the present invention, the construction of the fused tensor includes,

[0012] Obtain multi-dimensional runtime data to construct observation tensors;

[0013] Construct a multidimensional feature matrix based on the tensor rate of change;

[0014] The multidimensional feature matrix is ​​decomposed, and a fusion tensor is constructed based on the decomposition results.

[0015] As a preferred embodiment of the power grid fault analysis method of the present invention, the method includes: acquiring multi-dimensional operational data to construct the observation tensor,

[0016] Data cleaning and normalization are performed on multi-dimensional operational data acquired through multiple source sensors.

[0017] The cleaned and normalized multi-dimensional operational data are used to construct the observation tensor X according to the temporal, spatial, and feature dimensions. obs ;

[0018] A stop threshold and a preset algorithm are used to analyze the observed tensor X. obs Perform completion to obtain the completed tensor X, and set the constraint condition: subject to X. Ω =X obs , where X Ω It is a subset of the set of observation locations Ω.

[0019] In a preferred embodiment of the power grid fault analysis method of the present invention, the multidimensional feature matrix is ​​decomposed, and a fusion tensor is constructed based on the decomposition results, including:

[0020] The multidimensional feature matrix is ​​decomposed into a core tensor and a factor matrix using Tucker decomposition.

[0021] The network latency, bandwidth utilization, and packet loss rate data in the multi-dimensional operational data of the completed tensor X are synchronously sorted according to the time step to construct the communication network feature matrix C.

[0022] The communication network feature matrix C, the core tensor, and the factor matrix are combined using the tensor product formula to generate a fused tensor.

[0023] As a preferred embodiment of the power grid fault analysis method of the present invention, generating a fault report includes generating a fault report based on the results of power grid fault classification and power grid equipment fault areas.

[0024] As a preferred embodiment of the power grid fault analysis method of the present invention, the multi-element sensor includes voltage, current, temperature, frequency, bandwidth monitoring, delay monitoring and packet loss monitoring sensors.

[0025] Multi-dimensional operational data includes voltage, current, temperature, frequency, network latency, bandwidth utilization, and packet loss rate.

[0026] In a preferred embodiment of the power grid fault analysis method of the present invention, the preset criterion is a relative error stopping criterion;

[0027] The default algorithm is a high-precision low-rank tensor completion algorithm.

[0028] Secondly, the present invention provides a power grid fault analysis method, comprising: a construction generation module for constructing a fusion tensor and performing dimensionality reduction processing on the fusion tensor to generate a fusion feature matrix;

[0029] Construct a detection model to build an anomaly fusion feature matrix for an anomaly detection using a convolutional autoencoder model;

[0030] The calculation and generation module is used to classify power grid faults based on the abnormal data of the abnormal feature matrix, calculate the fault area of ​​power grid equipment based on the results of the power grid fault classification, and generate a fault report.

[0031] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0032] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method described above.

[0033] Compared with existing technologies, the beneficial effects of this invention are as follows: By employing a high-precision low-rank tensor completion algorithm to complete power grid operation data, the integrity and accuracy of the data are ensured. Utilizing Tucker decomposition and matrix compression techniques, key features are extracted from multi-dimensional data collected from multiple sources of sensors, and a fused feature matrix is ​​constructed, effectively improving data processing efficiency and feature representation capabilities. The application of a convolutional autoencoder model enhances the detection accuracy of abnormal features, while methods based on improved weighted summation and hierarchical clustering improve the accuracy of fault classification and location. Finally, fault reports are displayed in real time through a visual interface, enabling rapid identification, precise location, and intuitive display of power grid faults, significantly improving the intelligent monitoring level of the power grid and the response speed of fault handling. Attached Figure Description

[0034] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0035] Figure 1 This is a flowchart of a power grid fault analysis method.

[0036] Figure 2 This is a schematic diagram of the internal structure of a computer device. Detailed Implementation

[0037] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0038] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0039] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0040] Example 1

[0041] Reference Figure 1 This is the first embodiment of the present invention, which provides a power grid fault analysis method, comprising:

[0042] S1. Construct a fusion tensor and perform dimensionality reduction on the fusion tensor to generate a fusion feature matrix.

[0043] Furthermore, constructing the fusion tensor includes,

[0044] Obtain multi-dimensional runtime data to construct observation tensors.

[0045] Furthermore, acquiring multi-dimensional operational data to construct the observation tensor includes,

[0046] Data cleaning and normalization are performed on multi-dimensional operational data acquired through multiple source sensors.

[0047] The cleaned and normalized multi-dimensional operational data are used to construct the observation tensor X according to the temporal, spatial, and feature dimensions. obs ;

[0048] It should be noted that X obs ∈R I×J×T Where I is the number of sensor types, J is the number of spatial nodes, representing the physical location of the sensors, and T is the time length, representing the temporal information of the data, for each observation. Let R be the measurement value of the i-th type of sensor at the j-th node and time t, and R be the set of real numbers.

[0049] A stop threshold and a preset algorithm are used to analyze the observed tensor X. obs Perform completion to obtain the completed tensor X, and set the constraint condition: subject to X. Ω =X obs , where X Ω It is a subset of the set of observation locations Ω.

[0050] Furthermore, the preset criterion is the relative error stopping criterion;

[0051] The default algorithm is a high-precision low-rank tensor completion algorithm.

[0052] It should be noted that a stopping threshold is set using the relative error stopping criterion. A high-precision low-rank tensor completion algorithm is then used to complete the observed tensor Xobs where some values ​​are missing, resulting in the completed tensor X. The formula is as follows:

[0053]

[0054] In the formula, N represents the number of modes in the tensor, corresponding to the sensor type, spatial nodes, and temporal dimension, respectively, (X). n Let ||(X) be the expansion matrix of tensor X in the nth mode. n || * Let λ be the nuclear norm, representing the sum of the singular values ​​of the nth mode expansion matrix, used to preserve the low-rank property of the tensor. n Let X be the weight hyperparameter of the expansion matrix in the nth mode, controlling the degree of low rank of the expansion matrix in each mode; X is the completed tensor; and α is the regularization coefficient, controlling the strength of the retained constraint terms. Let be the matrix norm 2, representing the difference between the completed tensor X and the observed tensor X. obs The difference at observed locations, where Ω is the set of observed locations, and x i，j，t Let be the tensor value of the i-th type of sensor at the j-th node and time t. Let be the measurement value of the i-th type of sensor at the j-th node and time t.

[0055] It should be further noted that the observation tensor X is used. obs An initial tensor is constructed. For observation locations belonging to the observation location set, the measurement value of the i-th type of sensor at the j-th node and time t remains unchanged. For missing locations not belonging to the observation location set, the initial value is set to zero. The initialized tensor is expanded into a matrix form along different modes n, obtaining the expansion matrix for each mode. The purpose of tensor expansion is to convert the high-dimensional tensor into a matrix representation for subsequent calculation of the nuclear norm to perform low-rank constraints. To achieve high-precision low-rank tensor completion, an alternating direction optimization method is used to update the expansion matrix and elements at the missing locations for each mode in each iteration. On the expansion matrix of each mode, singular value decomposition is applied to orthogonal and diagonal matrices. Based on the nuclear norm terms in the objective function, a soft thresholding operation is applied to update the singular values. The expansion matrix is ​​reconstructed using the updated singular value matrix. The values ​​of the observation tensor are maintained on the set of observation locations, i.e., the constraints are satisfied. At the missing locations, the positions of the missing values ​​are re-estimated based on the update results of the current mode expansion matrix to minimize the overall error of the tensor. The error value between the current iteration and the previous iteration is calculated, and the change in the objective function is judged. If the change in the objective function is less than the set stopping threshold, the iteration is stopped, and the final completed tensor is obtained.

[0056] Ideally, a constraint should be set: subject to X Ω =X obs , where X Ω A subset of the set of observation locations Ω is a subset of the observation tensor X. obs A subset selection ensures that the completion result is consistent with the known data at the observation location by selecting only the values ​​at the observation location. This selection and constraint setting helps the algorithm focus on completing missing values ​​without changing the observed values, ensuring that the completed tensor conforms to the low-rank property and maintains the consistency of the observed data.

[0057] By maintaining the observed data values ​​at the observation locations, the completion results are ensured to be consistent with the actual observed values, preventing unnecessary changes to the known data and thus ensuring the authenticity of the completed data. This constraint limits the adjustment range of the optimization algorithm, allowing the optimization process to focus on the missing data region, thereby improving the completion efficiency. The constraint effectively prevents over-adjustment on the observed data and avoids the algorithm overfitting the observed data, thus ensuring that the completed tensor has good generalization ability.

[0058] The low-rank constraint of the tensor ensures that the completed tensor structure maintains the compactness and intrinsic correlation of multidimensional data from power grids and communication networks. There are often strong correlations between the multimodal features of power grid and communication data; for example, voltage and current data of specific spatial nodes at different time periods are highly correlated. Low-rank constraints can preserve these correlations. Furthermore, low-rank constraints effectively remove noise, as power grid and communication data often contain noise. The low-rank structure helps filter out irrelevant noisy data and maintain the integrity of key fault features. Compared to direct interpolation methods, low-rank constraints avoid local data bias, while other methods cannot fully capture the global features of multidimensional data, potentially leading to data inconsistency or over-smoothing when completing high-dimensional data. In power grid data, fault features often appear only at certain observation locations and manifest as abnormal voltage and current fluctuations. To avoid losing this crucial fault information, retention constraints must be used to ensure that the completed result is completely consistent with the observed data at known locations. Power grid observation data may contain key indicators of electrical equipment faults, such as voltage surges and frequency fluctuations. This information is crucial for operation and maintenance analysis. If retention constraints are not used during the completion process, the completed data will be incomplete. The complete result may deviate from the observed data, leading to distortion or loss of fault features. Retaining constraints ensures the accuracy of the observed data, guaranteeing that the completed result precisely matches the original data at known data points without introducing bias. In contrast, relying solely on low-rank constraints or other methods (such as least squares) may cause the completed result to deviate from the true data at known data points, resulting in inaccuracy. Traditional low-rank tensor completion only focuses on low-rank constraints while neglecting the retention of observed data, leading to bias at known data locations. This scheme adds retention constraints to ensure low-rank completion of data while preserving the accuracy of observed data. Power grid and communication network data are complex and multimodal. The improved optimization objective function, through low-rank constraints of the multimodal expansion matrix and the retention of observed data, better adapts to the characteristics of power grid fault data, avoiding the loss of key fault features due to smoothing during completion. Compared to other tensor completion methods (such as weighted interpolation and mean-filling), this method guarantees the accuracy and reliability of the completed data, providing higher-precision basic data support in power grid fault analysis, enabling the analysis model to more accurately identify and locate faults.

[0059] Furthermore, the multi-sensor includes sensors for voltage, current, temperature, frequency, bandwidth monitoring, delay monitoring, and packet loss monitoring;

[0060] Multi-dimensional operational data includes voltage, current, temperature, frequency, network latency, bandwidth utilization, and packet loss rate.

[0061] It should be noted that using multi-source sensors for data acquisition allows for comprehensive monitoring of various power grid operating parameters, covering important indicators such as voltage, current, and temperature. This diversity helps in better understanding the overall state of the power grid, thereby improving the accuracy of fault detection. Data cleaning and normalization are key steps in improving data quality. By cleaning the data, noise and inconsistencies can be eliminated, ensuring the reliability of the basic data for analysis. This process not only improves the integrity of the data but also provides clean input for subsequent analysis. Normalization ensures that data with different features are on the same order of magnitude, thus eliminating the adverse effects caused by differences in units, making model training more efficient. The construction of the observation tensor allows multi-dimensional data to be organized in a more efficient form. Facilitating subsequent analysis and calculation, the combination of temporal, spatial, and feature dimensions comprehensively reflects the operating status of the power grid, providing rich information for fault detection. The application of a high-precision low-rank tensor completion algorithm greatly enhances the system's ability to handle missing data. Through this algorithm, missing values ​​in the observation tensor can be effectively completed, ensuring the integrity and accuracy of the data. This completion process not only improves the availability of data but also optimizes the predictive performance of subsequent models. The use of a relative error stopping criterion ensures the efficiency of the completion algorithm. By dynamically adjusting the stopping threshold, the computation time can be reduced while ensuring completion accuracy, improving overall processing efficiency. This enables the system to respond quickly and adapt to different operating environments in practical applications.

[0062] Furthermore, a multidimensional feature matrix is ​​constructed based on the tensor rate of change.

[0063] It should be noted that the finite difference method was used to calculate the temporal and spatial rates of change R of the completed tensor X. T R S ,formula:

[0064]

[0065] In the formula, Δt is the time interval, Δ is the distance between adjacent nodes, and X is the distance between adjacent nodes. i,j,t Let X be the measurement value of the i-th type of sensor at the j-th node and time t;

[0066] The feature enhancement matrix construction method is used to construct the model based on the temporal and spatial rates of change R of the completed tensor X. T R S Construct a multidimensional feature matrix M.

[0067] Furthermore, the multidimensional feature matrix is ​​decomposed, and a fusion tensor is constructed based on the decomposition results.

[0068] Furthermore, the multidimensional feature matrix is ​​decomposed, and a fusion tensor is constructed based on the decomposition results, including...

[0069] The multidimensional feature matrix is ​​decomposed into a core tensor and a factor matrix using Tucker decomposition.

[0070] The network latency, bandwidth utilization, and packet loss rate data in the multi-dimensional operational data of the completed tensor X are synchronously sorted according to the time step to construct the communication network feature matrix C.

[0071] The communication network feature matrix C, the core tensor, and the factor matrix are combined using the tensor product formula to generate a fused tensor.

[0072] It should be noted that by performing a tensor product between the core tensor and the factor matrix, a new intermediate tensor is obtained. This intermediate tensor is then subjected to a tensor product with another factor matrix, and features are further merged. Finally, the merged features are combined with the last factor matrix to complete the tensor product of all modes, resulting in a fused tensor D, which contains all the important information of the power grid feature matrix and the communication network feature matrix.

[0073] It should be further explained that matrix compression technology is used to reduce the dimensionality of the fusion tensor D, generating the fusion feature matrix L.

[0074] Preferably, constructing a multidimensional feature matrix based on tensor change rates provides an efficient method for power grid fault analysis. Calculating the temporal and spatial change rates of the completed tensor helps to dynamically monitor the power grid's operating status. This process can promptly capture changes in the power grid and provide accurate fault early warning information. Through this dynamic monitoring, the power system can respond to potential faults more quickly, reduce the risk of power outages, and improve the overall reliability of the power grid. Using Tucker decomposition to analyze the multidimensional feature matrix can effectively extract the core features from the data. The process simplifies complex high-dimensional data into core tensors and factor matrices, making subsequent data processing more efficient. In the power system, it can quickly identify key factors affecting power grid stability, providing a basis for decision-making. The construction of fusion tensors enhances the system's comprehensive understanding of faults by integrating data from different sources, combining communication characteristics such as network latency, bandwidth utilization, and packet loss rate with power grid operating data. Combining these methods allows for a more accurate assessment of the power grid's operational status and its interaction with communication networks. This comprehensive analysis helps reveal the potential causes of faults, leading to more precise fault classification and location. The application of matrix compression technology further improves analysis efficiency. By reducing dimensionality, the complexity of the data is reduced, making subsequent model training and fault detection more efficient. Power grid systems typically contain massive amounts of data. Matrix compression technology can retain important information while reducing the demand for storage and computing resources, which is particularly important for real-time monitoring and fault response. Feature-enhanced matrix construction utilizes information on temporal and spatial rates of change to significantly improve the expressive power of the feature matrix. This method enhances sensitivity to minute changes, making fault detection more proactive. This is crucial for quickly identifying potential fault hazards during power grid operation, helping to take preventative measures in advance and reduce the probability of accidents.

[0075] S2. Construct a convolutional autoencoder model to detect anomalies and fuse feature matrices.

[0076] It should be noted that constructing the anomaly detection fusion feature matrix of the convolutional autoencoder model refers to obtaining multi-dimensional running data with normal labels from scientific datasets, preprocessing it to generate the model training set;

[0077] Construct a convolutional autoencoder model, including an encoder and a decoder;

[0078] Set the input format of the model to a fused feature matrix;

[0079] The convolutional autoencoder model is trained using the model training set, and the model parameters are iteratively optimized using the loss function and the Adam optimizer.

[0080] The fused feature matrix L is input into the trained convolutional autoencoder model to obtain the reconstructed feature matrix. The error feature matrix between the fused feature matrix and the reconstructed feature matrix is ​​calculated using the loss function.

[0081] Based on historical multi-dimensional operational data, a reconstruction error threshold is set. The error feature matrix is ​​compared with the reconstruction error threshold, and the element values ​​in the error feature matrix are filtered out to generate an abnormal feature matrix. If the element value is greater than the reconstruction error threshold, a 1 is marked at the corresponding position in the abnormal feature matrix; otherwise, a 0 is marked. The element value refers to the absolute value of the difference between each element in the input fused feature matrix and the reconstructed feature matrix.

[0082] Ideally, by constructing a convolutional autoencoder model and training it with normal labels from multi-dimensional operational data, the accuracy and sensitivity of power grid fault detection are greatly improved. Preprocessing the model training set by collecting normally labeled data from scientific datasets ensures that the features learned by the model represent normal operating patterns. This data-driven approach enables the model to quickly adapt and effectively identify potential anomalies when facing actual power grid operation conditions. The convolutional autoencoder design includes an encoder and a decoder. The encoder extracts data features through convolutional layers, reducing dimensionality while retaining key information. The decoder is responsible for reconstructing the input data. This structure effectively captures spatial relationships in the data and is particularly suitable for processing multi-dimensional power grid operational data. During training, the model parameters are iteratively optimized using a loss function and the Adam optimizer, which can quickly converge to the optimal solution, thereby improving the model's performance. This method, which enhances robustness, inputs the fused feature matrix into a trained convolutional autoencoder model to generate a reconstructed feature matrix. Then, the error between the original and reconstructed data is calculated using a loss function. This process accurately reveals the limitations of the model reconstruction and helps identify feature variations that should not exist under normal operating conditions. By setting a reconstruction error threshold, normal and abnormal feature matrices can be clearly distinguished, thus generating an abnormal feature matrix. This strategy enables fault detection to not only rely on historical data but also dynamically adapt to new operating states, improving the timeliness and accuracy of detection. In power grid management, the generation of abnormal feature matrices is a crucial basis for fault classification and location. By analyzing abnormal features, potential fault areas and their causes can be quickly identified, helping maintenance personnel take appropriate measures. This method not only improves the safety and stability of the power grid but also lays the foundation for intelligent management of the power system.

[0083] S3. Classify power grid faults based on the abnormal data in the abnormal feature matrix, calculate the fault areas of power grid equipment based on the results of the power grid fault classification, and generate a fault report.

[0084] It should be noted that classifying power grid faults based on the abnormal data in the abnormal feature matrix and calculating the fault area of ​​power grid equipment based on the results of power grid fault classification refers to scanning the position marked as 1 in the abnormal feature matrix and extracting the corresponding abnormal data from the fused feature matrix L.

[0085] Collect and preprocess multi-dimensional operational data of historical normal labels. Use statistical control limits to set normal ranges for the preprocessed multi-dimensional operational data of historical normal labels. Compare the extracted abnormal data with the normal range and retain the abnormal data that are greater than the normal range.

[0086] The retained outlier data is normalized, and K-Means clustering is used to classify the fault types of the normalized outlier data.

[0087] Using the locations of power grid equipment in the normalized outlier data as nodes, a topology is established based on the actual physical connections of the power grid, and an adjacency matrix A is defined, A∈R. N×N Where N is the total number of device nodes in the power grid, and the elements A of the adjacency matrix are... ij , indicating whether node i and node j are directly connected, and R is the set of real numbers;

[0088] The z-score is used to calculate the outlier f of node i in the k-th type of power grid fault. ik ;

[0089] The weighted summation formula is:

[0090]

[0091] In the formula, S 0,i To calculate the outlier score of node i using weighted summation, ω k f represents the weight of the k-th type of power grid fault, where K is the total number of power grid fault categories. ik Let be the outlier value of node i in the k-th type of power grid fault;

[0092] The outlier score S of node i is calculated using the improved weighted summation. i The formula is:

[0093]

[0094] In the formula, K represents the total number of power grid fault categories, and g(f ik ) is a nonlinear function, B is the total number of nodes in the power grid, h(f) is a nonlinear function. ik ) is the square root function, used to limit the excessive amplification of outliers from adjacent nodes on the current node, and k is the index of the fault type, used to distinguish different fault types;

[0095] Ideally, some isolated nodes exist in the power grid. Without adding an autocorrelation term, the adjacency term in the denominator becomes invalid when a node has no adjacent nodes, failing to reasonably reflect the node's anomalous score. The autocorrelation term ensures that even without adjacent nodes, each node can calculate its anomalous score based on its own anomalousness, thus guaranteeing the universality and stability of the formula. The adjacency matrix reflects the connection relationship between nodes, enabling the formula to spatially reflect the anomalous interactions between nodes according to the power grid topology. Using the adjacency matrix to extend outliers to the adjacent regions of nodes can effectively detect regional anomalies and improve the accuracy of power grid fault location. Without using the adjacency matrix, relying solely on the node's own anomalous values ​​cannot accurately capture regional faults, losing the ability to detect fault propagation and regional interactions. Outliers may vary drastically, and a single linear operation cannot balance the influence of various anomalous features on the node score. The nonlinear function not only amplifies the influence of high outliers but also controls the influence of outliers on the total score. The improved formula, by adding a node autocorrelation term to the denominator, allows the formula to operate normally even without adjacent nodes, making it applicable to various power grid topologies and improving the universality of the calculation. Compared with existing technologies, this formula incorporates the outlier values ​​of adjacent nodes into the calculation using an adjacency matrix, reflecting the spatial correlation of power grid nodes. In regional faults, the outlier values ​​of adjacent nodes will affect the score of the target node, thereby improving the detection capability of local anomalies. Traditional methods often ignore the influence of adjacent nodes and are difficult to capture regional fault characteristics. By smoothing and controlling the influence of extreme outliers, the formula ensures the stability of the calculation and improves the sensitivity to outlier scores.

[0096] It should be further explained that the nonlinear function g(f) ik The formula is:

[0097]

[0098] Square root function h(f) ik The formula is:

[0099]

[0100] Use the percentile method to set the outlier score threshold, and then compare the outlier score threshold with the outlier score S of node i. i Compare the results and retain the outlier score S of node i. i Nodes with scores exceeding the abnormal score threshold are marked as faulty nodes;

[0101] Hierarchical clustering is used to set a maximum path distance threshold. The connections between faulty nodes are found using an adjacency matrix. If A is satisfied... ik =1, then it is determined to be the same fault area. If it is A ik If the distance is 0, the Floyd-Warshall algorithm is used to calculate the shortest path distance between faulty nodes. The shortest path distance is compared with the maximum path distance threshold. If the shortest path distance is less than or equal to the maximum path distance threshold, they are set as the same fault area. If the shortest path distance is greater than the maximum path distance threshold, they are set as isolated fault areas.

[0102] By comparing the anomaly feature matrix with historical normal data, an efficient fault location and analysis method is provided. The extracted anomaly feature matrix identifies potential fault points, and the positions marked as 1 can quickly locate anomaly data in the fused feature matrix, laying the foundation for subsequent fault analysis. When preprocessing multi-dimensional operational data with historical normal labels, a statistical control limit method is used to set the normal range, ensuring that the data reflects the true operating state. This process not only improves the accuracy of anomaly data screening but also provides a reliable basis for subsequent analysis. After normalizing the retained anomaly data, the K-Means clustering algorithm is used to classify fault types. This method enables accurate identification of different fault types, facilitating subsequent analysis and response. Normalization helps eliminate the influence of different units of measurement, making data comparisons more reasonable. By constructing the power grid topology and defining the adjacency matrix A, the connection relationships between devices can be clearly reflected. This provides a crucial foundation for subsequent fault area identification, enabling the effective assessment of fault probability for each node through its anomaly score. Using z-score to calculate outliers for specific fault types, combined with an improved weighted summation method, yields a more accurate anomaly score. This score not only comprehensively considers the node's own anomaly but also appropriately accounts for the influence of adjacent nodes, effectively reducing the over-amplification effect of outliers. After setting an anomaly score threshold, comparison with the node's anomaly score allows for rapid identification of faulty nodes. This mechanism ensures the system can respond and maintain promptly when potential faults are detected, reducing the impact of faults on the power grid. By combining hierarchical clustering and the Floyd-Warshall algorithm with an adjacency matrix, the connection relationships between faulty nodes can be determined, clearly identifying the same fault area. This process makes fault location more precise, especially in complex power grid environments, helping maintenance personnel to quickly take measures to prevent fault escalation.

[0103] Furthermore, generating fault reports includes generating fault reports based on the results of power grid fault classification and power grid equipment fault areas. These fault reports can be displayed through a visual interface.

[0104] It should be noted that generating a fault report from the results of power grid fault classification and power grid equipment fault area includes obtaining the fault type of each node from the classification of node anomaly scores and fault types, sorting the fault nodes in the fault area from largest to smallest, setting the largest anomaly score of the fault node as the fault center node, generating a fault report, and converting the fault report into PDF format.

[0105] By obtaining the fault type of each node from the node anomaly score and fault type classification, the health status of power grid equipment can be systematically assessed. This process enables maintenance personnel to accurately identify which equipment is at risk of failure, thus providing a scientific basis for subsequent maintenance decisions. After obtaining the fault type of each node, sorting the faulty nodes in the fault area from largest to smallest helps to clearly show the severity of the fault. This sorting not only makes the faulty nodes readily apparent but also helps maintenance personnel quickly find the nodes that need to be prioritized. Setting the faulty node with the largest anomaly score as the fault center node can effectively focus on the most critical equipment, ensuring the rational allocation and efficient utilization of resources. This method helps to reduce fault response time, improve the overall stability of the power grid, and generate... The fault reporting process integrates all fault information in a structured manner, facilitating subsequent analysis and archiving. By providing reports containing key information such as fault type, fault node, and anomaly score, maintenance personnel can gain a more comprehensive understanding of the current operating status of the power grid. This report not only supports decision-making but also serves as a reference for daily maintenance, helping to develop targeted maintenance plans. Converting fault reports to PDF format ensures readability and ease of sharing. PDF documents can be easily viewed on different devices while maintaining their original format, avoiding information loss due to software differences. This shareable characteristic enables fault information to be quickly transmitted to managers and technical teams at all levels, promoting cross-departmental collaboration and improving response speed.

[0106] Furthermore, building a visual interface to display fault reports refers to using the Dash framework of Python to build a visual interface, using data visualization tools to display fault reports in real time, and using map visualization tools to display fault areas;

[0107] Users who have passed real-name verification are allowed to view this information.

[0108] The visualization interface built using the Dash framework can quickly respond to user input and provide an intuitive interactive experience. Developers can combine various chart types, such as line charts, bar charts, and heatmaps, to comprehensively display fault data. This not only improves user operability but also enhances data understandability, enabling technicians and management to quickly grasp the situation of power grid faults. Using data visualization tools to display fault reports in real time ensures that users obtain the latest information as soon as a fault occurs. This feature is crucial for the timely response of the power system, reducing delays in fault handling and improving the reliability and security of the power grid. By displaying fault areas through map visualization tools, users can intuitively understand the geographical distribution of faults and their impact on the power grid. This spatial analysis helps the operation and maintenance team quickly locate problem areas, effectively guiding on-site inspection and maintenance, and reducing the time and cost of fault recovery. Through real-name verification, the system ensures that only authorized users can view sensitive fault data and reports. This measure not only protects the company's trade secrets but also increases the security of data access, ensuring that the power company's operational information is not misused or leaked.

[0109] In summary, the beneficial effects of the power grid fault analysis method of this invention are as follows: by employing a high-precision low-rank tensor completion algorithm to complete power grid operation data, the integrity and accuracy of the data are ensured. Utilizing Tucker decomposition and matrix compression techniques, key features are extracted from multi-dimensional data collected from multiple sources of sensors, and a fused feature matrix is ​​constructed, effectively improving data processing efficiency and feature representation capabilities. The application of a convolutional autoencoder model enhances the detection accuracy of abnormal features, while the method based on improved weighted summation and hierarchical clustering improves the accuracy of fault classification and location. Finally, by displaying fault reports in real time through a visual interface, this technical solution achieves rapid identification, precise location, and intuitive display of power grid faults, significantly improving the intelligent monitoring level of the power grid and the response speed of fault handling.

[0110] Example 2

[0111] This embodiment provides a power grid fault analysis system, which includes a construction and generation module for constructing a fusion tensor and performing dimensionality reduction processing on the fusion tensor to generate a fusion feature matrix;

[0112] Construct a detection model to build an anomaly fusion feature matrix for an anomaly detection using a convolutional autoencoder model;

[0113] The calculation and generation module is used to classify power grid faults based on the abnormal data of the abnormal feature matrix, calculate the fault area of ​​power grid equipment based on the results of the power grid fault classification, and generate a fault report.

[0114] The above-mentioned unit modules can be embedded in the processor of the computer device in hardware form or independent of it, or they can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of the above modules.

[0115] Example 3

[0116] This embodiment provides a computer device, which may be a terminal, and its internal structure diagram may be as follows: Figure 2 As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a power grid fault analysis method. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0117] This embodiment also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the following: constructing a fusion tensor and performing dimensionality reduction processing on the fusion tensor to generate a fusion feature matrix; constructing a convolutional autoencoder model to detect abnormal fusion feature matrices; classifying power grid faults based on the abnormal data in the abnormal feature matrix; calculating the fault area of ​​power grid equipment based on the results of the power grid fault classification; and generating a fault report.

[0118] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A power grid fault analysis method, characterized in that: include, Constructing a fusion tensor and performing dimensionality reduction on the fusion tensor to generate a fusion feature matrix, wherein constructing the fusion tensor includes, Obtain multi-dimensional runtime data to construct observation tensors; Construct a multidimensional feature matrix based on the tensor rate of change; The multidimensional feature matrix is ​​decomposed, and the fusion tensor is constructed based on the decomposition results; The step of decomposing the multidimensional feature matrix and constructing the fusion tensor based on the decomposition result includes: The multidimensional feature matrix is ​​decomposed into a core tensor and a factor matrix using Tucker decomposition. The network latency, bandwidth utilization, and packet loss rate data in the multi-dimensional operational data of the completed tensor X are synchronously sorted according to the time step to construct the communication network feature matrix C. The communication network feature matrix C, the core tensor, and the factor matrix are combined using the tensor product formula to generate the fused tensor. Construct a convolutional autoencoder model to detect anomalies and fuse feature matrices; Among them, constructing a convolutional autoencoder model to detect anomalies and fusion feature matrix refers to obtaining multi-dimensional running data with normal labels from scientific datasets, preprocessing it to generate a model training set; This involves constructing a convolutional autoencoder model, which includes an encoder and a decoder; Set the input format of the model to a fused feature matrix; The convolutional autoencoder model is trained using the model training set, and the model parameters are iteratively optimized using the loss function and the Adam optimizer. The fused feature matrix L is input into the trained convolutional autoencoder model to obtain the reconstructed feature matrix. The error feature matrix between the fused feature matrix and the reconstructed feature matrix is ​​calculated using the loss function. Based on historical multi-dimensional operating data, a reconstruction error threshold is set. The error feature matrix is ​​compared with the reconstruction error threshold, and the element values ​​in the error feature matrix are filtered out to generate an abnormal feature matrix. If the element value is greater than the reconstruction error threshold, it is marked with 1 at the corresponding position of the abnormal feature matrix; otherwise, it is marked with 0. The element value refers to the absolute value of the difference between each element between the input fused feature matrix and the reconstruction feature matrix. Power grid faults are classified according to the abnormal data in the abnormal fusion feature matrix. Based on the results of the power grid fault classification, the fault area of ​​power grid equipment is calculated and a fault report is generated. Specifically, classifying power grid faults according to the abnormal data in the abnormal feature matrix and calculating the fault area of ​​power grid equipment based on the results of the power grid fault classification means scanning the position marked as 1 in the abnormal feature matrix and extracting the corresponding abnormal data from the fusion feature matrix L. Collect and preprocess multi-dimensional operational data of historical normal labels. Use statistical control limits to set normal ranges for the preprocessed multi-dimensional operational data of historical normal labels. Compare the extracted abnormal data with the normal range and retain the abnormal data that are greater than the normal range. The retained outlier data is normalized, and K-Means clustering is used to classify the fault types of the normalized outlier data.

2. The power grid fault analysis method as described in claim 1, characterized in that: The process of acquiring multi-dimensional operational data to construct the observation tensor includes, Data cleaning and normalization are performed on multi-dimensional operational data acquired through multiple source sensors. The cleaned and normalized multi-dimensional operational data are used to construct the observation tensor X according to the temporal, spatial, and feature dimensions. obs ; A stop threshold is set using preset criteria, and a preset algorithm is used to analyze the observed tensor X. obs Perform completion to obtain the completed tensor X, and set the constraint condition: subject to X. Ω =X obs , where X Ω It is a subset of the set of observation locations Ω.

3. The power grid fault analysis method as described in claim 2, characterized in that: The generation of fault reports includes generating fault reports based on the results of power grid fault classification and power grid equipment fault areas.

4. The power grid fault analysis method as described in claim 3, characterized in that: The multi-source sensors include sensors for voltage, current, temperature, frequency, bandwidth monitoring, delay monitoring, and packet loss monitoring; The multi-dimensional operational data includes voltage, current, temperature, frequency, network latency, bandwidth utilization, and packet loss rate data.

5. The power grid fault analysis method as described in any one of claims 2 to 4, characterized in that: The preset criterion is a relative error stopping criterion; The preset algorithm is a high-precision low-rank tensor completion algorithm.

6. A power grid fault analysis system, employing the method described in any one of claims 1-5, characterized in that, include: A construction and generation module is used to construct a fusion tensor and perform dimensionality reduction processing on the fusion tensor to generate a fusion feature matrix; Construct a detection model to build an anomaly fusion feature matrix for an anomaly detection using a convolutional autoencoder model; The calculation and generation module is used to classify power grid faults based on the abnormal data of the abnormal fusion feature matrix, calculate the fault area of ​​power grid equipment based on the result of the power grid fault classification, and generate a fault report.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images