Information comprehensive processing method and device based on power grid fault analysis, equipment and medium

Abstract

This invention discloses a method, apparatus, equipment, and medium for information integration processing based on power grid fault analysis, relating to the field of information integration processing technology for power grid fault analysis. The method includes: constructing a multi-source fault analysis model of the power grid based on pre-acquired multi-source fault feature data using a grey relational analysis algorithm; generating a data feature sequence based on the multi-source fault analysis model using a wavelet transform algorithm; generating a time-scaled calibration result based on the data feature sequence using a dynamic time warping algorithm; updating the multi-source fault feature data using a time-series interpolation algorithm based on the time-scaled calibration result to obtain a corrected fault feature sequence; and fusing the acquired power grid topology and the corrected fault feature sequence using a support vector machine algorithm to generate a fault judgment result. This application aims to solve the problems of low time-scaled calibration accuracy and lack of topology correlation in traditional methods.

Description

Technical Field

[0001] This invention relates to the field of information integration and processing technology for power grid fault analysis, and more specifically, to a method, apparatus, equipment, and medium for information integration and processing based on power grid fault analysis. Background Technology

[0002] As power systems develop towards higher voltage, larger capacity, and cross-regional interconnection, the power grid structure is becoming increasingly complex, and the operating environment is becoming more volatile. The suddenness, diversity, and cascading nature of faults pose severe challenges to the safe and stable operation of the power grid. Rapid and accurate fault analysis is the core of ensuring reliable power supply and shortening fault recovery time. This process highly depends on the effective processing and in-depth analysis of the massive amounts of fault characteristic data generated during power grid operation. Currently, power grid fault monitoring relies on a multi-source heterogeneous data system formed by various types of sensors and monitoring equipment. For example, SCADA systems collect macroscopic operating data, but the low sampling frequency makes it difficult to capture transient characteristics. DPMU devices collect transient data at high frequencies, which can reflect the dynamic evolution of faults. However, there are significant differences in the acquisition and transmission of multi-source data. Insufficient clock synchronization accuracy leads to time scale mismatch, which disrupts the correlation of characteristics. Communication interference and equipment failures cause data loss, seriously affecting the integrity and reliability of the data.

[0003] Existing power grid fault analysis methods have significant limitations in the comprehensive processing of multi-source data. Traditional methods often analyze single data sources or simply weight and fuse multi-source data, failing to fully consider the characteristic correlations and complementarities of different data types, making it difficult to construct a comprehensive fault feature model and resulting in biased analysis. Regarding the time-scale mismatch problem, existing methods often employ linear interpolation or fixed threshold calibration, ignoring the nonlinear dynamic characteristics of data time series during the fault process; calibration errors affect the accuracy of subsequent feature extraction. Fault features contain multi-scale information, but traditional time-domain or frequency-domain analysis struggles to separate components of different frequency bands, easily missing key transient features and reducing the discriminative power of feature sequences. Furthermore, existing models often use fault feature data in isolation, failing to fully integrate fault propagation path information dependent on topology, making it difficult to adapt to the needs of fault location and type identification under complex topologies, resulting in low accuracy.

[0004] For example, the invention patent application with publication number CN113761927B provides a method, system, device, and storage medium for real-time auxiliary decision-making in power grid fault handling. The method includes: acquiring fault information; searching and reasoning about fault information based on a power grid fault handling knowledge graph; and combining real-time power flow data of power grid equipment for real-time auxiliary decision-making. By constructing a power grid fault handling knowledge graph and combining it with prior knowledge of power grid dispatching, a real-time auxiliary decision-making method for power grid faults is designed, which can effectively provide decision-making support for power grid fault handling and support power grid dispatching work. Based on knowledge reasoning from the power grid fault handling knowledge graph and real-time power flow data of the power grid, combined with the actual situation of dispatching operations, the efficiency of power grid fault handling is improved, power grid power supply protection work is supported, and the intelligent development of power business is promoted.

[0005] For example, the invention patent publication number CN105224667B discloses a method for multi-station fault diagnosis and auxiliary decision-making based on intelligent power grid monitoring information. This method includes: establishing an expert database; decomposing multi-station power grid faults into single-station fault information; classifying fault information into switch quantity information reflecting the operating status of substation equipment and continuous change information of process data during the accident; describing switch quantity information using state elements; logically coupling and judging the state elements to determine the faulty component; performing time identification; performing spatial identification; performing feature quantity identification to determine the nature of the fault; matching with the expert database and performing coupled reasoning to find an auxiliary decision-making method; and displaying and tracking the data and decision results. This invention utilizes computers and monitoring systems to provide auxiliary decision-making for multi-station power grid faults, avoiding the drawbacks of manually processing massive amounts of monitoring data and insufficient personnel experience. It can quickly and accurately make decisions on multi-station power grid faults, preventing accidents from occurring.

[0006] The above-disclosed technical solutions have at least the following technical problems:

[0007] While existing solutions utilize knowledge graphs, real-time data, or expert databases to support fault-based decision-making, they fall short in the deep integration of multi-source heterogeneous fault feature data. They fail to fully exploit the inherent correlations between different types of monitoring data and lack a precise calibration mechanism for data timescale mismatch, making it difficult to ensure the consistency of fault features across time dimensions and potentially leading to biased decision-making. Furthermore, feature extraction in these solutions often remains at the surface level, failing to utilize effective multi-scale decomposition methods to mine key frequency band features during fault transients. The dynamic integration of power grid topology and fault features is also insufficient, relying more on experience bases than real-time integration of topological node connections. In complex power grid topologies, insufficient feature information mining or inadequate topological correlation can limit the accuracy of fault diagnosis and its adaptability to complex scenarios. These solutions fail to comprehensively address core issues such as multi-source data fusion, timescale calibration, deep feature extraction, and dynamic integration of topology and features.

[0008] To address the above problems, this invention proposes a solution. Summary of the Invention

[0009] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an information integration processing method, apparatus, equipment, and medium based on power grid fault analysis. By using wavelet transform and dynamic time warping to refine the time scale, and by using support vector machines to fuse power grid topology and fault features, the present invention solves the problems of low time scale calibration accuracy and lack of topology correlation in traditional methods.

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] The information integration processing method based on power grid fault analysis is characterized by the following steps: constructing a power grid multi-source fault analysis model based on pre-acquired multi-source fault feature data using a grey relational analysis algorithm; generating a data feature sequence using a wavelet transform algorithm based on the power grid multi-source fault analysis model; generating a time-scaled calibration result using a dynamic time warping algorithm based on the data feature sequence; updating the multi-source fault feature data using a time-series interpolation algorithm based on the time-scaled calibration result to obtain a corrected fault feature sequence; and fusing the acquired power grid topology and the corrected fault feature sequence using a support vector machine algorithm to generate a fault judgment result.

[0012] In a preferred embodiment, the multi-source fault characteristic data includes effective voltage values, effective current values, disconnector opening and closing status data, node voltage phase angle, current phasor amplitude, and current phase angle data.

[0013] In a preferred embodiment, the step of constructing a power grid multi-source fault analysis model based on pre-acquired multi-source fault feature data using a grey relational analysis algorithm specifically involves: analyzing the multi-source fault feature data using a grey relational analysis algorithm to establish feature mapping relationships; and fusing the feature mapping relationships using a weighted fusion algorithm to construct the power grid multi-source fault analysis model.

[0014] In a preferred embodiment, the step of generating a data feature sequence using a wavelet transform algorithm based on a power grid multi-source fault analysis model specifically involves: extracting first feature data of multi-source faults based on the power grid multi-source fault analysis model; performing multi-scale decomposition using a wavelet transform algorithm based on the first feature data of multi-source faults to obtain wavelet coefficient sequences of different frequency bands; and extracting time-domain and frequency-domain features based on the wavelet coefficient sequences to generate a data feature sequence.

[0015] In a preferred embodiment, the step of generating time-scaled calibration results based on the data feature sequence using a dynamic time warping algorithm specifically involves: constructing a distance matrix of the data feature sequence using an Euclidean distance algorithm based on the data feature sequence; and generating time-scaled calibration results based on the distance matrix using a dynamic time warping algorithm.

[0016] In a preferred embodiment, the step of updating the multi-source fault feature data to obtain the corrected fault feature sequence based on the time-stamped calibration results using a time-series interpolation algorithm specifically involves: identifying the missing moments and time-stamp mismatch intervals of the multi-source fault feature data on the time axis based on the time-stamped calibration results; performing numerical estimation on the mismatched or missing feature data points using a time-series interpolation algorithm based on the missing moments and time-stamp mismatch intervals to obtain numerical estimation results; and updating the time labels and corresponding feature values ​​of the multi-source fault feature data according to the numerical estimation results to obtain the corrected fault feature sequence.

[0017] In a preferred embodiment, the step of fusing the acquired power grid topology and the corrected fault feature sequence using a support vector machine algorithm to generate a fault judgment result specifically involves: extracting node connection relationships from the power grid topology using graph theory modeling methods to obtain a topology feature set; normalizing the corrected fault feature sequence using a Min-Max normalization algorithm to obtain a normalized fault feature set; constructing the input feature matrix of the support vector machine using a feature concatenation algorithm based on the topology feature set and the normalized fault feature set; and using the input feature matrix as input to generate a fault judgment result through the support vector machine algorithm.

[0018] An information processing device based on power grid fault analysis is characterized by comprising a feature modeling module, a sequence extraction module, a time-scale calibration module, a feature correction module, and a fault judgment module, with interconnections between the modules: the feature modeling module constructs a power grid multi-source fault analysis model based on pre-acquired multi-source fault feature data using a grey relational analysis algorithm; the sequence extraction module generates a data feature sequence using a wavelet transform algorithm based on the power grid multi-source fault analysis model; the time-scale calibration module generates a time-scale calibration result based on the data feature sequence using a dynamic time warping algorithm; the feature correction module updates the multi-source fault feature data to obtain a corrected fault feature sequence based on the time-scale calibration result using a time-series interpolation algorithm; and the fault judgment module fuses the acquired power grid topology and the corrected fault feature sequence using a support vector machine algorithm to generate a fault judgment result.

[0019] An electronic device, characterized in that the electronic device comprises: at least one processor; and an input / output interface communicatively connected to the at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform an information integration processing method based on power grid fault analysis.

[0020] A computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements an information integration processing method based on power grid fault analysis.

[0021] The technical effects and advantages of the information integration processing method, device, equipment, and medium based on power grid fault analysis of this invention are as follows:

[0022] This invention proposes an information integration processing method, device, equipment, and medium based on power grid fault analysis. It utilizes a grey relational analysis algorithm to mine the inherent correlations of multi-source fault feature data, and combines this with weighted fusion to construct an analysis model. This overcomes the limitations of traditional single data sources or simple weighted fusion, fully leveraging the complementarity of different data types to improve the comprehensiveness of fault feature characterization. Simultaneously, it employs a wavelet transform algorithm to perform multi-scale decomposition of multi-source fault data, accurately extracting time-domain and frequency-domain features, effectively capturing key frequency band information during fault transient processes, and addressing the problems of shallow feature extraction and omission of key transient features in traditional analysis methods, thus enhancing the discriminative power of feature sequences. A distance matrix is ​​constructed using a dynamic time warping algorithm to achieve nonlinear time-scale calibration. Combined with time-series interpolation, missing or mismatched data are numerically estimated, correcting time labels and feature values, overcoming the error problems of traditional linear interpolation or fixed threshold calibration, and ensuring the consistency and integrity of data in the time dimension. Finally, graph theory modeling is used to extract the connection relationships of power grid topology nodes, and support vector machines are used to fuse topological structure features with corrected fault features, deeply associating fault propagation paths with feature data, addressing the shortcomings of existing models that use feature data in isolation and lack sufficient topological correlation. It effectively solves the problems of low time-scale calibration accuracy and lack of topological correlation in traditional methods. Attached Figure Description

[0023] Figure 1 This is a schematic flowchart of an information integration and processing method based on power grid fault analysis provided in an embodiment of the present invention;

[0024] Figure 2 This is a schematic diagram of the information integration and processing device based on power grid fault analysis provided in an embodiment of the present invention;

[0025] Figure 3 This is a schematic diagram of the information integration and processing equipment based on power grid fault analysis provided in an embodiment of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0028] Example 1, Figure 1 This invention presents an information integration and processing method based on power grid fault analysis, characterized by the following steps:

[0029] S1, Based on the pre-acquired multi-source fault characteristic data, a power grid multi-source fault analysis model is constructed using the grey relational analysis algorithm;

[0030] S2, Based on the power grid multi-source fault analysis model, a data feature sequence is generated using the wavelet transform algorithm;

[0031] S3, Based on the data feature sequence, the time-scale calibration result is generated through a dynamic time warping algorithm;

[0032] S4. Based on the time-stamped calibration results, the multi-source fault feature data is updated using a time-series interpolation algorithm to obtain the corrected fault feature sequence;

[0033] S5 integrates the acquired power grid topology and corrected fault feature sequences using a support vector machine algorithm to generate fault judgment results.

[0034] This invention proposes an information integration processing method, device, equipment, and medium based on power grid fault analysis. It utilizes a grey relational analysis algorithm to mine the inherent correlations of multi-source fault feature data, and combines this with weighted fusion to construct an analysis model. This overcomes the limitations of traditional single data sources or simple weighted fusion, fully leveraging the complementarity of different data types to improve the comprehensiveness of fault feature characterization. Simultaneously, it employs a wavelet transform algorithm to perform multi-scale decomposition of multi-source fault data, accurately extracting time-domain and frequency-domain features, effectively capturing key frequency band information during fault transient processes, and addressing the problems of shallow feature extraction and omission of key transient features in traditional analysis methods, thus enhancing the discriminative power of feature sequences. A distance matrix is ​​constructed using a dynamic time warping algorithm to achieve nonlinear time-scale calibration. Combined with time-series interpolation, missing or mismatched data are numerically estimated, correcting time labels and feature values, overcoming the error problems of traditional linear interpolation or fixed threshold calibration, and ensuring the consistency and integrity of data in the time dimension. Finally, graph theory modeling is used to extract the connection relationships of power grid topology nodes, and support vector machines are used to fuse topological structure features with corrected fault features, deeply associating fault propagation paths with feature data, addressing the shortcomings of existing models that use feature data in isolation and lack sufficient topological correlation. It effectively solves the problems of low time-scale calibration accuracy and lack of topological correlation in traditional methods.

[0035] S1. Based on the pre-acquired multi-source fault characteristic data, a multi-source fault analysis model for the power grid is constructed using the grey relational analysis algorithm.

[0036] In this embodiment, the multi-source fault characteristic data includes the effective value of voltage, the effective value of current, the opening and closing status data of disconnectors, the phase angle of node voltage, the amplitude of current phasors, and the phase angle data of current.

[0037] In this embodiment, the construction of a power grid multi-source fault analysis model based on pre-acquired multi-source fault feature data using a grey relational analysis algorithm specifically involves:

[0038] The feature mapping relationship is established by analyzing multi-source fault characteristic data using the grey relational analysis algorithm.

[0039] A multi-source fault analysis model for power grids is constructed by fusing feature mapping relationships using a weighted fusion algorithm.

[0040] Establishing feature mapping relationships using grey relational analysis is the core step in model construction. Multi-source fault feature data mainly includes steady-state measurement data (such as RMS voltage and current values) provided by the SCADA system and transient measurement data (such as current phase angle, voltage phase angle, and phasor amplitude) captured by the DPMU device. These data sources differ significantly in sampling frequency, temporal resolution, and feature dimensions: SCADA data is typically refreshed at the second or minute level, reflecting the macroscopic steady-state operation of the system; while DPMU can achieve millisecond or even microsecond-level sampling, accurately capturing subtle changes during transient processes. Grey relational analysis can effectively handle the correlation problem between such heterogeneous and non-uniform sequences. It quantitatively assesses the inherent coupling relationship between features by calculating the geometric similarity and consistency of change trends between different data sequences.

[0041] For example, when a short-circuit fault occurs in the power grid, the SCADA system records a sudden increase in the effective value of the current, while the DPMU can acquire a sharp jump in the current phase angle. The grey relational analysis algorithm can identify potential synchronous change patterns from these two types of data from different sources and at different scales, and assign them a quantified correlation strength. The specific algorithm steps are as follows:

[0042] Steady-state measurement data (SCADA): RMS voltage values, RMS current values, and disconnector status data. The steady-state measurement data refresh cycle is on the order of seconds or minutes. It is extended to a time axis consistent with the DPMU data using time interpolation or hold-and-hold methods. Transient measurement data (DPMU): Node voltage phase angle, current phasor amplitude, and current phase angle. The data sampling frequency is on the order of milliseconds or higher, and it is directly used as a high-resolution sequence. To ensure consistency of the grey relational analysis input, the two types of data need to be aligned: using the DPMU high-resolution time axis as a reference, the SCADA data is extended through linear interpolation.

[0043] If the effective value of the current is selected as the core fault characteristic, then the reference sequence Data in The effective value of the current in the th The original measurements at each time point, and the reference sequence. The formula is:

[0044]

[0045] Each feature forms a separate time series, denoted as:

[0046]

[0047] in, For the first Comparison sequences In the The value at each moment. For characteristic number, This represents the unified number of sampling times.

[0048] The formula for the grey relational analysis algorithm is:

[0049]

[0050] in, For at any time The correlation coefficient, Reference sequence In the The value at each moment. For the first Comparison sequences In the The value at each moment. This is the preset power grid short-circuit fault discrimination coefficient, with a default value of 0.5.

[0051] The feature mapping relationships are integrated using a weighted fusion algorithm. Based on the correlation coefficients obtained from grey relational analysis, dynamic weights are assigned to different feature mapping relationships: mapping relationships with higher correlation coefficients occupy higher weights in the model to ensure that critical features sensitive to faults are given sufficient attention; while secondary mapping relationships with lower correlation coefficients are given lower weights to avoid irrelevant information interfering with the model's accuracy.

[0052] Weighted fusion is technically achieved through matrix operations, integrating multiple sets of scattered feature mapping relationships into a unified model framework. Specifically, it includes the following steps: First, a weight matrix is ​​constructed based on the correlation coefficient matrix, where each weight coefficient reflects the relative importance of the corresponding feature pair; then, the original feature data and the weight matrix are weighted and synthesized, for example, using linear weighting or rule-based feature aggregation, ultimately forming a new fused feature set. The weighted fusion process integrates multiple sets of feature mapping relationships into a unified model framework through matrix operations, preserving the macroscopic descriptive ability of SCADA data for the overall operation of the power grid while incorporating the precise characterization of transient details during fault events by DPMU data. Ultimately, it constructs a multi-source fault analysis model for the power grid covering steady-state and transient, macroscopic and microscopic aspects, providing a structured and highly correlated data foundation for subsequent feature extraction, effectively overcoming the feature fragmentation problem caused by traditional single data sources or simple splicing methods.

[0053] S2, based on the power grid multi-source fault analysis model, generates a data feature sequence through wavelet transform algorithm.

[0054] In this embodiment, the step of generating a data feature sequence using a wavelet transform algorithm based on the power grid multi-source fault analysis model specifically involves:

[0055] Based on the power grid multi-source fault analysis model, the first feature data of multi-source faults are extracted;

[0056] Based on the first feature data of multi-source faults, multi-scale decomposition is performed using wavelet transform algorithm to obtain wavelet coefficient sequences of different frequency bands;

[0057] Based on the wavelet coefficient sequence, time-domain features and frequency-domain features are extracted to generate a data feature sequence.

[0058] The first feature data of multi-source faults is extracted based on the multi-source fault analysis model of the power grid. This model has integrated the multi-source data correlation between SCADA and DPMU through grey relational analysis and weighted fusion. Therefore, the extracted first feature data not only covers steady-state characteristics such as the effective voltage and current values ​​of the SCADA system, but also includes transient characteristics such as the voltage phase angle and current phasor amplitude of the DPMU device. Moreover, these features have been integrated through the model framework to ensure that the extracted first data is both comprehensive and has internal logical consistency, avoiding the feature isolation problem caused by traditional single data source extraction.

[0059] Multi-scale decomposition of fault features based on multi-source fault first feature data is performed using wavelet transform. The wavelet transform algorithm possesses time-frequency localization analysis capabilities, enabling precise decomposition of the non-stationary characteristics of fault signals. Specifically, the algorithm selects appropriate wavelet basis functions to perform multi-level decomposition of the first feature data sequence: sequentially decomposing it from low frequency to high frequency to obtain wavelet coefficient sequences in different frequency bands. Low-frequency coefficients correspond to the steady-state trend characteristics during the fault process, while high-frequency coefficients reflect the transient impact characteristics at the moment of the fault. This multi-scale decomposition overcomes the limitations of traditional single-domain or frequency-domain analysis, achieving refined decomposition of fault features across different time scales and frequency dimensions.

[0060] The wavelet transform algorithm formula is:

[0061]

[0062]

[0063] in, For the first Approximate coefficients of fault characteristics obtained after layer decomposition For the first Approximate coefficients of fault characteristics obtained from layer decomposition. This is the impulse response of the low-pass filter. This represents the impulse response of the high-pass filter. For the first The fault feature detail coefficients obtained after layer decomposition.

[0064] This paper utilizes wavelet coefficient sequences to extract time-domain and frequency-domain features, generating a data feature sequence. In time-domain feature extraction, statistics such as peak value, mean, variance, and kurtosis are calculated for wavelet coefficient sequences in different frequency bands to capture the amplitude variation patterns of the signal over time. For example, the peak value of high-frequency coefficients can reflect the intensity of transient fault impacts, while the variance of low-frequency coefficients can reflect the degree of fluctuation in steady-state characteristics. In frequency-domain feature extraction, Fourier transform or power spectrum analysis is performed on the wavelet coefficient sequences to obtain features such as the energy proportion and center frequency of each frequency band, quantifying the distribution characteristics of the fault signal at different frequency components. For example, grounding faults may exhibit significant energy accumulation in specific low-frequency bands. The extracted time-domain and frequency-domain features are integrated to form a data feature sequence containing multi-dimensional and multi-scale information. This sequence preserves the dynamic evolution details of the fault and enhances the discriminative power of the features through frequency domain analysis, effectively solving the problems of missed transient features and single feature dimensions in traditional feature extraction.

[0065] S3 generates time-scale calibration results based on the data feature sequence using a dynamic time warping algorithm.

[0066] In this embodiment, the step of generating time-scale calibration results based on the data feature sequence using a dynamic time warping algorithm specifically involves:

[0067] Based on the data feature sequence, a distance matrix of the data feature sequence is constructed using the Euclidean distance algorithm;

[0068] Time-scale calibration results are generated based on the distance matrix using a dynamic time warping algorithm.

[0069] A distance matrix is ​​constructed based on the data feature sequences using the Euclidean distance algorithm. The data feature sequences extracted in step S2 contain time and frequency domain features across multiple frequency bands. However, these features originate from different devices such as SCADA and DPMU, and may exhibit time axis misalignment due to asynchronous sampling clocks. For example, the occurrence time of the same fault event recorded in SCADA data may differ from the transient response time in DPMU data. The Euclidean distance algorithm quantifies the numerical differences in features by calculating the spatial distance between corresponding data points in different feature sequences. For two feature sequences to be calibrated, the algorithm treats the feature value at each time point in the sequence as a point in a multidimensional space, calculates the Euclidean distance between any two points, and ultimately forms a two-dimensional distance matrix. Each element in the matrix represents the degree of difference between the features at two time points in the two sequences, visually presenting the misalignment of the feature sequences in the time-numerical dimension.

[0070] Time-scaled calibration results are generated based on a distance matrix using a dynamic time warping algorithm. This algorithm addresses the calibration problem of sequences with non-uniform lengths and misaligned time scales by finding the optimal path to achieve non-linear alignment. Specifically, the algorithm uses a distance matrix as its foundation, searching for a path with the minimum cumulative distance from the top left to the bottom right corner. Each node in the path corresponds to a pair of time points in two sequences that need alignment. During the search, the algorithm uses constraints (such as maximum time offset) to prevent excessive path distortion, ensuring that the calibration results conform to the physical laws of power grid fault evolution. Ultimately, the time offset relationship corresponding to this optimal path is the time-scaled calibration result, clearly indicating the amount of time deviation that needs to be adjusted for each feature sequence. This achieves consistent alignment of multi-source feature data on the time axis, laying the foundation for subsequent data filling and time-scale error correction, and effectively overcoming the limitation of traditional linear interpolation calibration in handling non-linear time misalignments.

[0071] S4. Based on the time-stamped calibration results, the multi-source fault feature data is updated using a time-series interpolation algorithm to obtain the corrected fault feature sequence.

[0072] In this embodiment, the step of updating the multi-source fault feature data based on the time-stamped calibration results to obtain the corrected fault feature sequence using a time-series interpolation algorithm specifically involves:

[0073] Based on the time-scale calibration results, identify the missing moments and time-scale mismatch intervals of multi-source fault characteristic data on the time axis;

[0074] Based on the missing time and time mismatch interval, the feature data points that are mismatched or missing are numerically estimated using a time-series interpolation algorithm to obtain the numerical estimation result.

[0075] Based on the numerical estimation results, the time labels and corresponding feature values ​​of the multi-source fault feature data are updated to obtain the corrected fault feature sequence.

[0076] Data problems can be identified based on time-stamped calibration results. The time-stamped calibration results clearly define the misalignment relationships of multi-source feature sequences on the time axis. Based on this, two types of problems in the data can be accurately located: first, missing moments, i.e., blank points in feature data acquisition caused by communication interruptions, equipment failures, etc. For example, a DPMU device may experience a missing phase angle data at a certain millisecond time point due to data transmission delay at the moment of a sudden fault; second, time-stamp mismatch intervals, referring to the deviation range of feature sequences from different data sources on the time stamp. For example, there is a systematic deviation interval of 0.3 seconds between the time stamp of the disconnector opening / closing status recorded by the SCADA system and the corresponding voltage change time stamp of the DPMU. By traversing the time axis and combining the time offset in the calibration results, these problem points and intervals can be clearly marked, providing a clear target for subsequent corrections.

[0077] Numerical estimation is performed based on the identified missing time points and time-scale mismatch intervals. For different data problems, the time-series interpolation algorithm employs differentiated estimation strategies: for isolated missing data points, linear interpolation or spline interpolation methods are used to fit reasonable values ​​based on the valid data before and after the missing point—for example, if the effective voltage value at a certain moment is missing, the missing value can be calculated using linear trends based on the effective values ​​one second before and one second after the missing point; for continuous data within the time-scale mismatch interval, numerical redistribution is achieved through time axis mapping. For example, transient characteristic data leading 0.5 seconds in the DPMU sequence is shifted to the corresponding SCADA time axis position according to the calibration results, and the gap data generated during the shift is supplemented by interpolation. The algorithm fully considers the continuity and smoothness characteristics of power grid faults during estimation, avoiding abrupt numerical changes that do not conform to physical laws, and ensuring the reasonableness of the estimation results.

[0078] The data is updated based on numerical estimation results, and a corrected sequence is generated. The time labels and corresponding feature values ​​of multi-source fault characteristic data are updated bidirectionally: on the one hand, the time labels of all data within the mismatch interval are uniformly adjusted to the calibrated reference time axis, eliminating time deviations from different data sources; on the other hand, the interpolated estimated values ​​are used to fill in the feature gaps at missing moments, replacing unreasonable values ​​within the mismatch interval. After the update, all feature data are strictly aligned in the time dimension, and there are no data breaks or contradictions, ultimately forming a complete and consistent corrected fault characteristic sequence. This sequence retains the effective information from the initial data and eliminates the interference of time label misalignment and data loss through correction, providing high-quality input data for subsequent fault judgment based on the power grid topology, effectively improving the accuracy and reliability of fault analysis.

[0079] The formula for generating the corrected sequence is:

[0080]

[0081] in, To calculate the target time Multi-source fault characteristic estimation value, For the target time of the feature sequence that needs interpolation, For the target time The timestamp of the most recent valid data point in the feature sequence. For the target time The timestamp of the most recent valid data point in the feature sequence. for The effective eigenvalues ​​corresponding to time 1. for The effective feature value corresponding to the time step.

[0082] S5 integrates the acquired power grid topology and corrected fault feature sequences using a support vector machine algorithm to generate fault judgment results.

[0083] In this embodiment, the process of fusing the acquired power grid topology and the corrected fault feature sequence using a support vector machine algorithm to generate a fault judgment result is as follows:

[0084] The node connection relationships of the power grid topology are extracted using graph theory modeling methods to obtain the topology feature set;

[0085] The corrected fault feature sequence is normalized using the Min-Max normalization algorithm to obtain a normalized fault feature set.

[0086] Based on the topological feature set and the normalized fault feature set, the input feature matrix of the support vector machine is constructed through the feature concatenation algorithm;

[0087] The input feature matrix is ​​used as input, and the fault judgment result is generated through the support vector machine algorithm.

[0088] Graph theory modeling is used to extract the topological feature set of a power grid. The power grid topology reflects the physical connections between nodes and is a crucial carrier of fault propagation paths. Graph theory modeling abstracts the power grid into a "node-edge" model: substations, line endpoints, etc., are considered nodes, and line connections are considered edges. Topological features are quantified using mathematical forms such as adjacency matrices and degree matrices—for example, in the adjacency matrix, "1" indicates two nodes are directly connected, and "0" indicates no direct connection; the degree matrix records the number of connections for each node, reflecting the importance of each node in the power grid. Furthermore, features such as shortest path length and connected components can be extracted to form a topological feature set, providing spatial correlation information of the power grid structure for subsequent integration.

[0089] The fault feature sequence is processed using the Min-Max normalization algorithm. This sequence contains multi-dimensional features such as effective values, phase angles, and amplitudes, with significant differences in the numerical ranges of different features. Min-Max normalization maps all feature values ​​to the [0,1] interval, eliminating the influence of dimensional differences on the model—for example, it normalizes the effective current value from "500A~2000A" to a value in the "0-1" range, ensuring that features such as voltage phase angle and current amplitude have equal weight in model training, and avoiding model bias towards certain features due to differences in numerical ranges. The normalized fault feature set obtained after normalization preserves the distribution pattern of the features while improving the convergence speed and stability of the algorithm.

[0090] The support vector machine input matrix is ​​constructed based on two types of feature sets through feature concatenation. The feature concatenation algorithm concatenates the topological feature set and the normalized fault feature set according to the sample dimension: for each fault sample, its corresponding topological feature vector is concatenated with the normalized fault feature vector end to end to form a high-dimensional input feature matrix. For example, the topological feature vector of a sample is [0.2, 0.5, 0.3], and the normalized fault feature vector is [0.1, 0.8, 0.4, 0.6]. After concatenation, an input vector of [0.2, 0.5, 0.3, 0.1, 0.8, 0.4, 0.6] is formed. Each row in the matrix corresponds to the comprehensive feature of a sample, realizing the organic integration of "spatial topology - temporal features".

[0091] The input feature matrix is ​​fed into a Support Vector Machine (SVM) algorithm to generate fault diagnosis results. SVM identifies fault types or locations by finding the optimal classification hyperplane, making it particularly suitable for classification problems in high-dimensional feature spaces. The algorithm first maps the input features to a higher-dimensional space to address nonlinear classification; then, it determines support vectors by maximizing the classification margin, constructing a classification model; finally, it substitutes the input feature matrix of the sample to be diagnosed into the model, outputting fault type labels or fault node coordinates, etc. Because the input features integrate the spatial correlation of the power grid topology with the temporal dynamics of fault features, the model can fully utilize the physical law that "fault propagation paths are topologically constrained." For example, when a line fault occurs, the voltage characteristic changes of its associated nodes will highly match the topological adjacency relationship, thus significantly improving the accuracy of fault diagnosis and effectively solving the location bias problem caused by the isolated use of feature data in traditional models.

[0092] The decision function formula for support vector machines is:

[0093]

[0094] in, For the input feature vector, The normal vector of the hyperplane is the input feature matrix. For bias terms, The optimal classification hyperplane for the input feature matrix. For the decision outcome.

[0095] Based on test data from 500 sets of power grid fault cases, a comparison of key technologies between the traditional method and the method of this invention was obtained, as shown in Table 1:

[0096] Table 1. Comparison of key technologies between traditional methods and the method of this invention.

[0097]

[0098] Example 2, Figure 2This invention discloses an information integration and processing device based on power grid fault analysis, characterized by comprising a feature modeling module, a sequence extraction module, a time-scale calibration module, a feature correction module, and a fault judgment module, with interconnections between the modules: The feature modeling module constructs a power grid multi-source fault analysis model based on pre-acquired multi-source fault feature data using a grey relational analysis algorithm; the sequence extraction module generates a data feature sequence using a wavelet transform algorithm based on the power grid multi-source fault analysis model; the time-scale calibration module generates a time-scale calibration result based on the data feature sequence using a dynamic time warping algorithm; the feature correction module updates the multi-source fault feature data to obtain a corrected fault feature sequence based on the time-scale calibration result using a time-series interpolation algorithm; and the fault judgment module fuses the acquired power grid topology and the corrected fault feature sequence using a support vector machine algorithm to generate a fault judgment result.

[0099] Example 3, Figure 3 The present invention discloses an information integration processing device based on power grid fault analysis, comprising: at least one processor; and an input / output interface communicatively connected to the at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform an information integration processing method based on power grid fault analysis.

[0100] A computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements an information integration processing method based on power grid fault analysis.

[0101] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0102] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0103] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0104] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0105] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0106] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An information integration and processing method based on power grid fault analysis, characterized in that, Includes the following steps: Based on pre-acquired multi-source fault feature data, a multi-source fault analysis model for power grid is constructed using a grey relational analysis algorithm. Based on the power grid multi-source fault analysis model, a data feature sequence is generated using the wavelet transform algorithm; Based on the data feature sequence, time-scaled calibration results are generated using a dynamic time warping algorithm. Based on the time-stamped calibration results, the corrected fault feature sequence is obtained by updating the multi-source fault feature data through a time-series interpolation algorithm; The acquired power grid topology and the corrected fault feature sequence are fused using a support vector machine algorithm to generate fault judgment results.

2. The information integration and processing method based on power grid fault analysis according to claim 1, characterized in that, The multi-source fault characteristic data includes effective voltage values, effective current values, disconnector opening and closing status data, node voltage phase angle, current phasor amplitude, and current phase angle data.

3. The information integration and processing method based on power grid fault analysis according to claim 2, characterized in that, The process of constructing a power grid multi-source fault analysis model based on pre-acquired multi-source fault feature data using a grey relational analysis algorithm is as follows: The feature mapping relationship is established by analyzing multi-source fault characteristic data using the grey relational analysis algorithm. A multi-source fault analysis model for power grids is constructed by fusing feature mapping relationships using a weighted fusion algorithm.

4. The information integration and processing method based on power grid fault analysis according to claim 3, characterized in that, The process of generating data feature sequences using wavelet transform algorithm based on the power grid multi-source fault analysis model is as follows: Based on the power grid multi-source fault analysis model, the first feature data of multi-source faults are extracted; Based on the first feature data of multi-source faults, multi-scale decomposition is performed using wavelet transform algorithm to obtain wavelet coefficient sequences of different frequency bands; Based on the wavelet coefficient sequence, time-domain features and frequency-domain features are extracted to generate a data feature sequence.

5. The information integration and processing method based on power grid fault analysis according to claim 4, characterized in that, The process of generating time-scale calibration results based on the data feature sequence using a dynamic time warping algorithm is as follows: Based on the data feature sequence, a distance matrix of the data feature sequence is constructed using the Euclidean distance algorithm; Time-scale calibration results are generated based on the distance matrix using a dynamic time warping algorithm.

6. The information integration and processing method based on power grid fault analysis according to claim 5, characterized in that, The process of updating multi-source fault feature data based on time-stamped calibration results to obtain a corrected fault feature sequence using a time-series interpolation algorithm is as follows: Based on the time-scale calibration results, identify the missing moments and time-scale mismatch intervals of multi-source fault characteristic data on the time axis; Based on the missing time and time mismatch interval, the feature data points that are mismatched or missing are numerically estimated using a time-series interpolation algorithm to obtain the numerical estimation result. Based on the numerical estimation results, the time labels and corresponding feature values ​​of the multi-source fault feature data are updated to obtain the corrected fault feature sequence.

7. The information integration and processing method based on power grid fault analysis according to claim 6, characterized in that, The process of fusing the acquired power grid topology and corrected fault feature sequence using a support vector machine algorithm to generate fault judgment results is as follows: The node connection relationships of the power grid topology are extracted using graph theory modeling methods to obtain the topology feature set; The corrected fault feature sequence is normalized using the Min-Max normalization algorithm to obtain a normalized fault feature set. Based on the topological feature set and the normalized fault feature set, the input feature matrix of the support vector machine is constructed through the feature concatenation algorithm; The input feature matrix is ​​used as input, and the fault judgment result is generated through the support vector machine algorithm.

8. An apparatus for using the information integration processing method based on power grid fault analysis as described in any one of claims 1-7, characterized in that, Includes the following modules: The feature modeling module constructs a power grid multi-source fault analysis model based on pre-acquired multi-source fault feature data and uses a grey relational analysis algorithm. The sequence extraction module generates data feature sequences based on the power grid multi-source fault analysis model using wavelet transform algorithm; The time-scale calibration module generates time-scale calibration results based on the data feature sequence using a dynamic time warping algorithm; The feature correction module, based on the time-stamped calibration results, updates the multi-source fault feature data using a time-series interpolation algorithm to obtain a corrected fault feature sequence; The fault diagnosis module fuses the acquired power grid topology and corrected fault feature sequences using a support vector machine algorithm to generate fault diagnosis results.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and an input / output interface communicatively connected to the at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the information integration processing method based on power grid fault analysis as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the information integration processing method based on power grid fault analysis as described in any one of claims 1-7.

Images