Intelligent fault identification system and method for power supply and distribution system

Abstract

The invention relates to the technical field of power distribution system monitoring and fault identification, in particular to an intelligent fault identification system and method for a power supply and distribution system. The method comprises the following steps: acquiring conventional operation data of a target, abnormal operation data before a fault occurs, fault survey data and fault processing scheme data; determining a secondary label system and a multilevel system severity system of historical faults according to the fault survey data; screening out electrical internal and cross-type strong correlation feature pairs; constructing a feature pool, determining an association combination of a plurality of features and a main fault tag for the feature pool and the secondary tag system, and determining an association score of the association combination; and monitoring the abnormal features, determining early warning fault labels of the abnormal features, and performing early warning based on the early warning fault label types. According to the method, strong correlation feature pairs are accurately screened, the stability is analyzed, contribution mutual information and conditional entropy are combined to reduce redundancy, the relevance between features and faults is obtained, and the accuracy of fault recognition is improved.

Description

Technical Field

[0001] This invention relates to the field of power distribution system monitoring and fault identification technology, specifically to an intelligent fault identification system and method for power supply and distribution systems. Background Technology

[0002] The power supply and distribution system is the core of electricity supply. Fault identification can promptly detect anomalies, locate the cause of faults, reduce downtime and losses, and ensure stable power supply. Its importance lies in maintaining the normal operation of industrial production and residential life, avoiding large-scale power outages, and is a key guarantee for the safe and efficient operation of the power system.

[0003] In the field of power supply and distribution system fault identification, existing technical methods have the following shortcomings: feature correlation analysis is one-sided, often ignoring the strong cross-type correlation between electrical and environmental features; fault classification is coarse, lacking a refined labeling system; early warning relies on a single feature, resulting in low accuracy; fault handling solutions have poor matching degree and lack closed-loop management. These problems lead to delayed fault identification, high false alarm rate, increased maintenance costs and system downtime risks, and make it difficult to meet the needs of complex power supply and distribution systems for rapid fault location, accurate early warning and efficient handling. More comprehensive technical solutions are urgently needed to break through the bottlenecks. Summary of the Invention

[0004] The purpose of this invention is to address the problems existing in the background technology by proposing an intelligent fault identification system and method for power supply and distribution systems.

[0005] The technical solution of this invention: A method for intelligent fault identification in a power supply and distribution system, comprising the following steps: S1. Mark the power supply and distribution system as the target, and obtain the target's normal operation data, abnormal operation data before the fault occurred, fault investigation data, and fault handling plan data; S2. Determine the two-level labeling system and multi-level system severity system for historical faults based on fault investigation data; S3. Extract electrical features and physical and environmental features from abnormal operation data and normalize them. Combine the normalized features into feature pairs, calculate the Pearson correlation coefficient, and screen out strongly correlated feature pairs within and across electrical types. S4. Perform variance correlation calculation on strongly correlated feature pairs to determine stable feature pairs and construct a feature pool. Determine the association combinations of multiple features with the main fault label for the feature pool and the secondary label system. Perform correlation analysis on the association combinations and determine the correlation score of the association combinations. S5. Monitor abnormal features, determine the early warning fault labels of abnormal features based on the correlation scores of all associated combinations, and issue early warnings based on the types of early warning fault labels. S6. Obtain fault handling plan data based on the warning content and execute SOP trigger; record all linkage operations in detail during the fault handling process to form a complete fault handling plan, and track and evaluate the entire fault handling process to achieve closed-loop management.

[0006] Preferably, the two-level labeling system includes a main fault label and sub-fault labels; the main fault label includes multiple sub-fault labels.

[0007] Preferably, S3 includes: S31. Calculate the Pearson correlation coefficient of the feature pairs based on the following formula: ; In the formula, represents the Pearson correlation coefficient of the feature pair (Ti, Tj), used to quantify the degree of linear correlation between the two features; Ti(t) and Tj(t) are the sampled observations of features Ti and Tj at time t, respectively; μi and μj are the means of features Ti and Tj, respectively; i and j are the feature numbers; n is the total number of sampled samples; S32. Perform discrimination classification and threshold filtering on the feature pairs to obtain feature pairs containing electrical features. The type of the feature pairs is filtered using the following expression: ; In the formula, TL represents the feature pair type; The number of electrical features in a feature pair is obtained based on statistics; Indicates electrical internal characteristic pairs; Indicates cross-type feature pairs; Indicates other feature pairs; S33. Select electrical internal feature pairs and cross-type feature pairs as target feature pairs, compare them with the preset Pearson threshold, retain feature pairs with Pearson correlation coefficients greater than the Pearson threshold and mark them as strongly correlated feature pairs.

[0008] Preferably, S4 includes: S41. Calculate the variance of the Pearson correlation coefficient of strongly correlated feature pairs using a sliding window, and calculate the mean of the variances of the Pearson correlation coefficients of all windows to obtain the mean variance of strongly correlated feature pairs and compare it with the variance threshold. Strongly correlated feature pairs whose variance means are greater than a preset variance threshold are marked as stable feature pairs; S42. Construct a feature pool based on the features contained in several stable feature pairs. For each feature in the feature pool, obtain the normal parameter range of the feature based on normal operating data, and calculate the abnormal parameter range based on abnormal operating data and the normal parameter range.

[0009] Preferably, S4 also includes: S43. Randomly select any feature and any main fault label from the feature pool, and determine various anomaly definition conditions based on multiple sub-fault labels under the main fault label and the multi-level system severity. S44. Discretize the range of abnormal parameters of the features to obtain several discrete intervals of the features. S45. Construct a two-dimensional joint probability distribution matrix based on the feature discrete intervals and the anomaly definition conditions. The matrix dimension is [number of feature discrete intervals, anomaly definition conditions].

[0010] Preferably, in step S46, the joint probability matrix is ​​summed row by row to obtain the marginal probability P(X) of each feature interval, where X is the feature; the joint probability matrix is ​​summed column by column to obtain the marginal probability P(Y) of each anomaly definition condition, where Y is the system anomaly based on the main fault label. S47. The contribution value of the region using the following formula for the two-dimensional joint probability distribution matrix: ; In the formula, C(x, y) is the contribution value of the region (x, y) of the two-dimensional joint probability distribution matrix; P(X=x, Y=y) represents the joint probability that the feature variable takes the interval x and the anomaly definition condition y; P(X=x) represents the marginal probability that the feature variable is in the interval x; P(Y=y) represents the marginal probability that the anomaly definition condition y is y; x and y are the discrete interval number and the anomaly definition condition number, respectively, and both x and y are positive integers; S48. Sum the contribution values ​​of all regions in the two-dimensional joint probability distribution matrix to obtain the mutual information value of the contribution of feature X and system anomaly Y under the main fault label, and normalize the mutual information value.

[0011] Preferably, S4 also includes: S49. Calculate the conditional entropy of features Xh and Xk respectively, under the condition of obtaining the system anomaly Y under a given primary fault label, and label them as follows: and h and k are both feature numbers, h≠k; S410. Calculate the joint conditional entropy of features Xh and Xk under the given system anomaly Y with a primary fault label. ; The conditional mutual information values ​​of system anomaly Y, feature Xh, and feature Xk under a given primary fault label are calculated using the following formula. ; ; S411. Perform data analysis on the conditional mutual information values, and execute a redundancy processing strategy based on the analysis results, including processing the conditional mutual information values ​​of features Xh and Xk. Compared with a preset feature redundancy threshold, if the conditional mutual information value If the mutual information value is greater than the feature redundancy threshold, then features Xh and Xk are determined to have strong redundancy, and a redundancy processing strategy is executed; if the conditional ... If the value is not greater than the feature redundancy threshold, the redundancy processing strategy will not be executed. Based on the redundancy processing strategy, the mutual information values ​​of features Xh and Xk are compared, and the feature with the larger mutual information value is retained and determined as the conditional mutual information value. The results of data analysis.

[0012] Preferably, S4 also includes: S412. Obtain the analysis results obtained from data analysis of each mutual information value under different conditions, perform a union operation on the analysis results to obtain the feature union; sort the features in the feature union in descending order according to the contribution mutual information value to obtain the feature ranking, and obtain the feature ranking percentage value. Through formula The correlation score of the feature under the main fault label is calculated; where PB is the ranking percentage value; GV is the contribution mutual information value; α1 and α2 are the preset weights; S413. Repeat steps S43-S412 above to obtain the correlation score of each association combination. Divide the association combinations according to the features contained in the association combinations and determine the set of association combinations containing the same feature. S414. In the set of associated combinations, the associated combinations are sorted and filtered in descending order according to their correlation scores. The associated combinations with correlation scores greater than the preset score threshold are retained. The top m associated combinations are marked as target combinations. The main fault labels in the target combinations are combined to obtain the risk fault label set of the features.

[0013] Preferably, S5 includes: S51. When an anomaly is detected, obtain the corresponding risk fault label set; merge the risk fault label sets of the anomaly features to obtain the label statistics set; S52. Calculate the occurrence count of each main fault label in the label statistics set, mark the main fault label with the most occurrence count as the warning label, and issue a warning based on the fault type of the warning label.

[0014] This invention also discloses an intelligent fault identification system for power supply and distribution systems, which applies the above-mentioned intelligent fault identification method for power supply and distribution systems, specifically including: The data acquisition module is used to mark the power supply and distribution system as a target and acquire the target's normal operating data, abnormal operating data before the fault occurs, fault investigation data, and fault handling plan data. The tagging system construction module is used to determine the two-level tagging system and multi-level system severity system for historical faults based on fault investigation data; The data processing module is used to extract electrical features and physical and environmental features from abnormal operation data and normalize them. It combines the normalized features into feature pairs, calculates the Pearson correlation coefficient, and filters out strongly correlated feature pairs within and across electrical types. The fault identification and early warning module is used to monitor abnormal features, determine the early warning fault labels of abnormal features based on the correlation scores of all associated combinations, and issue early warnings based on the type of early warning fault labels. The SOP triggering and recording closed-loop management module is used to obtain fault handling plan data based on the warning content, execute SOP triggering, record all linkage operations in detail during the fault handling process to form a complete fault handling plan, and track and evaluate the entire fault handling process to achieve closed-loop management.

[0015] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects: (1) Achieve refined classification of faults through a two-level label system and a multi-level severity system; accurately screen strongly correlated feature pairs and analyze their stability; combine contribution mutual information and conditional entropy to reduce redundancy; obtain the correlation between features and faults; and improve the accuracy of fault identification. (2) Risk label set is generated based on correlation score to improve the accuracy of early warning; by dynamically matching fault handling schemes and closed-loop management, the processing efficiency is optimized, which solves the problems of insufficient feature correlation analysis, rough fault classification, low early warning accuracy and poor matching degree of handling scheme in the existing methods. Attached Figure Description

[0016] Figure 1 This is a layer structure diagram of Embodiment 1 of the present invention. Detailed Implementation

[0017] Example 1, as Figure 1 As shown, the present invention proposes an intelligent fault identification method for power supply and distribution systems, comprising the following steps: S1. Mark the power supply and distribution system as the target, and obtain the target's normal operation data, abnormal operation data before the fault occurred, fault investigation data, and fault handling plan data; S2. Determine the two-level labeling system and multi-level system severity system for historical faults based on fault investigation data. The two-level labeling system includes main fault labels and sub-fault labels. The main fault label includes multiple sub-fault labels. For example, the main fault labels include short-circuit faults, grounding faults, etc. Among them, grounding faults include metallic grounding faults, non-metallic grounding faults, permanent grounding faults, and transient grounding faults. The multi-level system severity system includes multiple levels of severity. For example, level one represents the most severe, level two represents moderately severe, and level three represents generally severe. It should be noted that the system severity is calculated by weighting the maintenance time, maintenance cost, and impact range of the power supply and distribution system.

[0018] S3. Extract electrical features and physical and environmental features from abnormal operation data and normalize them. Combine the normalized features into pairs, calculate the Pearson correlation coefficient, and screen out strongly correlated feature pairs within and across electrical types. For example, electrical features include waveform factor, peak factor, impulse factor, kurtosis factor, margin factor, skewness factor, waveform exponent, average amplitude, and RMS value, etc.; physical and environmental features include temperature value, temperature rise slope, humidity value, and amplitude, etc. S31. Calculate the Pearson correlation coefficient of the feature pairs based on the following formula: ; In the formula, represents the Pearson correlation coefficient of the feature pair (Ti, Tj), used to quantify the degree of linear correlation between the two features; Ti(t) and Tj(t) are the sampled observations of features Ti and Tj at time t, respectively; μi and μj are the means of features Ti and Tj, respectively; i and j are the feature numbers; n is the total number of sampled samples; S32. Perform discrimination classification and threshold filtering on the feature pairs to obtain feature pairs containing electrical features. The type of the feature pairs is filtered using the following expression: ; In the formula, TL represents the feature pair type; The number of electrical features in a feature pair is obtained based on statistics; Indicates electrical internal characteristic pairs; Indicates cross-type feature pairs; Indicates other feature pairs; S33. Select electrical internal feature pairs and cross-type feature pairs as target feature pairs, compare them with the preset Pearson threshold, retain feature pairs with Pearson correlation coefficients greater than the Pearson threshold and mark them as strongly correlated feature pairs.

[0019] S4. Perform variance correlation calculation on strongly correlated feature pairs to determine stable feature pairs and construct a feature pool. Determine the association combinations of multiple features with the main fault label for the feature pool and the secondary label system. Perform correlation analysis on the association combinations and determine the correlation score of the association combinations. S41. Calculate the variance of the Pearson correlation coefficient of strongly correlated feature pairs using a sliding window, and calculate the mean of the variances of the Pearson correlation coefficients of all windows to obtain the mean variance of strongly correlated feature pairs and compare it with the variance threshold. Strongly correlated feature pairs whose mean variance is greater than a preset variance threshold are marked as stable feature pairs. The variance threshold is calculated based on large datasets of the mean variance of strongly correlated feature pairs. A variance threshold higher than the threshold indicates that the Pearson correlation coefficient of the feature pair within the window remains highly consistent, maintaining a high degree of correlation and reflecting the stability of the correlation. S42. Construct a feature pool based on the features contained in several stable feature pairs. For each feature in the feature pool, obtain the normal parameter range of the feature based on normal operating data, and calculate the abnormal parameter range based on abnormal operating data and the normal parameter range. S43. Randomly select any feature and any main fault label from the feature pool, and determine various anomaly definition conditions based on multiple sub-fault labels under the main fault label and the multi-level system severity. S44. Discretize the range of abnormal parameters of the features to obtain several discrete intervals of the features. S45. Construct a two-dimensional joint probability distribution matrix based on the feature discrete intervals and the anomaly definition conditions. The matrix dimension is [number of feature discrete intervals, anomaly definition conditions]; S46. Sum the joint probability matrix by rows to obtain the marginal probability P(X) of each feature interval, where X is the feature; sum the joint probability matrix by columns to obtain the marginal probability P(Y) of each anomaly definition condition, where Y is the system anomaly based on the main fault label. S47. The contribution value of the region using the following formula for the two-dimensional joint probability distribution matrix: ; In the formula, C(x, y) is the contribution value of the region (x, y) of the two-dimensional joint probability distribution matrix; P(X=x, Y=y) represents the joint probability that the feature variable takes the interval x and the anomaly definition condition y; P(X=x) represents the marginal probability that the feature variable is in the interval x; P(Y=y) represents the marginal probability that the anomaly definition condition y is y; x and y are the discrete interval number and the anomaly definition condition number, respectively, and both x and y are positive integers; S48. Sum the contribution values ​​of all regions in the two-dimensional joint probability distribution matrix to obtain the mutual information value of the contribution between feature X and system anomaly Y under the main fault label, and normalize the mutual information value. It should be noted that the mutual information value only reflects the degree of correlation rather than causal relationship. S49. Calculate the conditional entropy of features Xh and Xk respectively, under the condition of obtaining the system anomaly Y under a given primary fault label, and label them as follows: and h and k are both feature numbers, h≠k; S410. Calculate the joint conditional entropy of features Xh and Xk under the given system anomaly Y with a primary fault label. It should be noted that the joint conditional entropy is calculated based on existing technical methods, which will not be elaborated on here. The conditional mutual information values ​​of system anomaly Y, feature Xh, and feature Xk under a given primary fault label are calculated using the following formula. ; ; S411. Perform data analysis on the conditional mutual information values, and execute a redundancy processing strategy based on the analysis results, including processing the conditional mutual information values ​​of features Xh and Xk. Compared with a preset feature redundancy threshold, if the conditional mutual information value If the mutual information value is greater than the feature redundancy threshold, then features Xh and Xk are determined to have strong redundancy, and a redundancy processing strategy is executed; if the conditional ... If the value is not greater than the feature redundancy threshold, the redundancy processing strategy will not be executed. Based on the redundancy processing strategy, the mutual information values ​​of features Xh and Xk are compared, and the feature with the larger mutual information value is retained and determined as the conditional mutual information value. The results of data analysis; S412. Obtain the analysis results obtained from data analysis of each mutual information value under different conditions, perform a union operation on the analysis results to obtain the feature union; sort the features in the feature union in descending order according to the contribution mutual information value to obtain the feature ranking, and obtain the feature ranking percentage value. Through formula The correlation score of the feature under the main fault label is calculated; where PB is the ranking percentage value; GV is the contribution mutual information value; α1 and α2 are the preset weights; S413. Repeat steps S43-S412 above to obtain the correlation score of each association combination. Divide the association combinations according to the features contained in the association combinations and determine the set of association combinations containing the same feature. S414. In the set of associated combinations, the associated combinations are sorted and filtered in descending order according to their correlation scores. The associated combinations with correlation scores greater than the preset score threshold are retained. The top m associated combinations are marked as target combinations. The main fault labels in the target combinations are combined to obtain the risk fault label set of the features.

[0020] S5. Monitor abnormal features, determine the early warning fault labels of abnormal features based on the correlation scores of all associated combinations, and issue early warnings based on the types of early warning fault labels. S51. When an anomaly is detected, obtain the corresponding risk fault label set; merge the risk fault label sets of the anomaly features to obtain the label statistics set; S52. Calculate the occurrence count of each main fault label in the label statistics set, mark the main fault label with the most occurrence count as the warning label, and issue a warning based on the fault type of the warning label.

[0021] S6. Obtain fault handling plan data based on the warning content and execute SOP trigger; record all linkage operations in detail during the fault handling process to form a complete fault handling plan, and track and evaluate the entire fault handling process to achieve closed-loop management; Specifically, the fault handling plan data includes several key pieces of information, including fault type, personnel involved in the handling, handling measures, system severity, percentage of recovery scope, and handling time; The system automatically records all key information throughout the entire process, from fault occurrence, warning issuance, SOP triggering, equipment linkage to fault troubleshooting; The matching value (Value) of the fault handling solution data is obtained by quantifying the system severity, recovery scope, and processing time. The quantification formula is as follows: In the formula, βz is the system severity quantification coefficient; SYz is the system severity; z is the system severity number, and z is a positive integer; Ct is the processing time. The matching values ​​(Value) of the fault handling plan data are sorted in descending order, and the fault handling plan data corresponding to the largest matching value (Value) is executed; in particular, a re-evaluation is performed after each fault handling; A refined fault classification is achieved through a two-level labeling system and a multi-level severity system; highly correlated feature pairs are accurately selected and their stability is analyzed; redundancy is reduced by combining contribution mutual information and conditional entropy; the correlation between features and faults is obtained, thereby improving the accuracy of fault identification. Risk label sets are generated based on correlation scoring to improve the accuracy of early warnings; processing efficiency is optimized by dynamically matching fault handling schemes and closed-loop management, which solves the problems of insufficient feature correlation analysis, coarse fault classification, low early warning accuracy and poor matching degree of handling schemes in existing methods.

[0022] Example 2: The intelligent fault identification system for power supply and distribution systems proposed in this invention is applied to the intelligent fault identification method for power supply and distribution systems proposed in Example 1, specifically including: The data acquisition module is used to mark the power supply and distribution system as a target and acquire the target's normal operating data, abnormal operating data before the fault occurs, fault investigation data, and fault handling plan data. The tagging system construction module is used to determine the two-level tagging system and multi-level system severity system for historical faults based on fault investigation data; The data processing module is used to extract electrical features and physical and environmental features from abnormal operation data and normalize them. It combines the normalized features into feature pairs, calculates the Pearson correlation coefficient, and filters out strongly correlated feature pairs within and across electrical types. The fault identification and early warning module is used to monitor abnormal features, determine the early warning fault labels of abnormal features based on the correlation scores of all associated combinations, and issue early warnings based on the type of early warning fault labels. The SOP triggering and recording closed-loop management module is used to obtain fault handling plan data based on the warning content, execute SOP triggering, record all linkage operations in detail during the fault handling process to form a complete fault handling plan, and track and evaluate the entire fault handling process to achieve closed-loop management.

[0023] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A method for intelligent fault identification in a power supply and distribution system, characterized in that, Includes the following steps: S1. Mark the power supply and distribution system as the target, and obtain the target's normal operation data, abnormal operation data before the fault occurred, fault investigation data, and fault handling plan data; S2. Determine the two-level labeling system and multi-level system severity system for historical faults based on fault investigation data; S3. Extract electrical features and physical and environmental features from abnormal operation data and normalize them. Combine the normalized features into feature pairs, calculate the Pearson correlation coefficient, and screen out strongly correlated feature pairs within and across electrical types. S4. Perform variance correlation calculation on strongly correlated feature pairs to determine stable feature pairs and construct a feature pool. Determine the association combinations of multiple features with the main fault label for the feature pool and the secondary label system. Perform correlation analysis on the association combinations and determine the correlation score of the association combinations. S5. Monitor abnormal features, determine the early warning fault labels of abnormal features based on the correlation scores of all associated combinations, and issue early warnings based on the types of early warning fault labels. S6. Obtain fault handling plan data based on the warning content and execute the SOP trigger; All coordinated operations during the fault handling process are recorded in detail to form a complete fault handling plan. The entire fault handling process is tracked and evaluated to achieve closed-loop management.

2. The intelligent fault identification method for a power supply and distribution system according to claim 1, characterized in that, The two-level labeling system includes main fault labels and sub-fault labels; the main fault label includes multiple sub-fault labels.

3. The intelligent fault identification method for a power supply and distribution system according to claim 2, characterized in that, S3 include: S31. Calculate the Pearson correlation coefficient of the feature pairs based on the following formula: ； In the formula, represents the Pearson correlation coefficient of the feature pair (Ti, Tj), used to quantify the degree of linear correlation between the two features; Ti(t) and Tj(t) are the sampled observations of features Ti and Tj at time t, respectively; μi and μj are the means of features Ti and Tj, respectively; i and j are the feature numbers; n is the total number of sampled samples; S32. Perform discrimination classification and threshold filtering on the feature pairs to obtain feature pairs containing electrical features. The type of the feature pairs is filtered using the following expression: ； In the formula, TL represents the feature pair type; The number of electrical features in a feature pair is obtained based on statistics; Indicates electrical internal characteristic pairs; Indicates cross-type feature pairs; Indicates other feature pairs; S33. Select electrical internal feature pairs and cross-type feature pairs as target feature pairs, compare them with the preset Pearson threshold, retain feature pairs with Pearson correlation coefficients greater than the Pearson threshold and mark them as strongly correlated feature pairs.

4. The intelligent fault identification method for a power supply and distribution system according to claim 3, characterized in that, S4 include: S41. Calculate the variance of the Pearson correlation coefficient of strongly correlated feature pairs using a sliding window, and calculate the mean of the variances of the Pearson correlation coefficients of all windows to obtain the mean variance of strongly correlated feature pairs and compare it with the variance threshold. Strongly correlated feature pairs whose variance means are greater than a preset variance threshold are marked as stable feature pairs; S42. Construct a feature pool based on the features contained in several stable feature pairs. For each feature in the feature pool, obtain the normal parameter range of the feature based on normal operating data, and calculate the abnormal parameter range based on abnormal operating data and the normal parameter range.

5. The intelligent fault identification method for a power supply and distribution system according to claim 4, characterized in that, S4 also includes: S43. Randomly select any feature and any main fault label from the feature pool, and determine various anomaly definition conditions based on multiple sub-fault labels under the main fault label and the multi-level system severity. S44. Discretize the range of abnormal parameters of the features to obtain several discrete intervals of the features. S45. Construct a two-dimensional joint probability distribution matrix based on the feature discrete intervals and the anomaly definition conditions. The matrix dimension is [number of feature discrete intervals, anomaly definition conditions].

6. The intelligent fault identification method for a power supply and distribution system according to claim 5, characterized in that, S4 also includes: S46. Sum the joint probability matrix by rows to obtain the marginal probability P(X) of each feature interval, where X is the feature; sum the joint probability matrix by columns to obtain the marginal probability P(Y) of each anomaly definition condition, where Y is the system anomaly based on the main fault label. S47. The contribution value of the region using the following formula for the two-dimensional joint probability distribution matrix: ； In the formula, C(x, y) is the contribution value of the region (x, y) of the two-dimensional joint probability distribution matrix; P(X=x, Y=y) represents the joint probability that the feature variable takes the interval x and the anomaly definition condition y; P(X=x) represents the marginal probability that the feature variable is in the interval x; P(Y=y) represents the marginal probability that the anomaly definition condition y is y; x and y are the discrete interval number and the anomaly definition condition number, respectively, and both x and y are positive integers; S48. Sum the contribution values ​​of all regions in the two-dimensional joint probability distribution matrix to obtain the mutual information value of the contribution of feature X and system anomaly Y under the main fault label, and normalize the mutual information value.

7. The intelligent fault identification method for a power supply and distribution system according to claim 5, characterized in that, S4 also includes: S49. Calculate the conditional entropy of features Xh and Xk respectively, under the condition of obtaining the system anomaly Y under a given primary fault label, and label them as follows: and h and k are both feature numbers, h≠k; S410. Calculate the joint conditional entropy of features Xh and Xk under the given system anomaly Y with a primary fault label. ; The conditional mutual information values ​​of system anomaly Y, feature Xh, and feature Xk under a given primary fault label are calculated using the following formula. ; ； S411. Perform data analysis on the conditional mutual information values, and execute a redundancy processing strategy based on the analysis results, including processing the conditional mutual information values ​​of features Xh and Xk. Compared with a preset feature redundancy threshold, if the conditional mutual information value If the mutual information value is greater than the feature redundancy threshold, then features Xh and Xk are determined to have strong redundancy, and a redundancy processing strategy is executed; if the conditional ... If the value is not greater than the feature redundancy threshold, the redundancy processing strategy will not be executed. Based on the redundancy processing strategy, the mutual information values ​​of features Xh and Xk are compared, and the feature with the larger mutual information value is retained and determined as the conditional mutual information value. The results of data analysis.

8. The intelligent fault identification method for a power supply and distribution system according to claim 7, characterized in that, S4 also includes: S412. Obtain the analysis results obtained from data analysis of each mutual information value under different conditions, perform a union operation on the analysis results to obtain the feature union; sort the features in the feature union in descending order according to the contribution mutual information value to obtain the feature ranking, and obtain the feature ranking percentage value. Through formula The correlation score of the feature under the main fault label is calculated; where PB is the ranking percentage value; GV is the contribution mutual information value; α1 and α2 are the preset weights; S413. Repeat steps S43-S412 above to obtain the correlation score of each association combination. Divide the association combinations according to the features contained in the association combinations and determine the set of association combinations containing the same feature. S414. In the set of associated combinations, the associated combinations are sorted and filtered in descending order according to their correlation scores. The associated combinations with correlation scores greater than the preset score threshold are retained. The top m associated combinations are marked as target combinations. The main fault labels in the target combinations are combined to obtain the risk fault label set of the features.

9. The intelligent fault identification method for a power supply and distribution system according to claim 8, characterized in that, S5 include: S51. When an anomaly is detected, obtain the corresponding risk fault label set; The risk fault label set with abnormal characteristics is merged to obtain the label statistics set; S52. Calculate the occurrence count of each main fault label in the label statistics set, mark the main fault label with the most occurrence count as the warning label, and issue a warning based on the fault type of the warning label.

10. A power supply and distribution system intelligent fault identification system, applied to the power supply and distribution system intelligent fault identification method according to any one of claims 1 to 9, characterized in that, Specifically, it includes: The data acquisition module is used to mark the power supply and distribution system as a target and acquire the target's normal operating data, abnormal operating data before the fault occurs, fault investigation data, and fault handling plan data. The tagging system construction module is used to determine the two-level tagging system and multi-level system severity system for historical faults based on fault investigation data; The data processing module is used to extract electrical features and physical and environmental features from abnormal operation data and normalize them. It combines the normalized features into feature pairs, calculates the Pearson correlation coefficient, and filters out strongly correlated feature pairs within and across electrical types. The fault identification and early warning module is used to monitor abnormal features, determine the early warning fault labels of abnormal features based on the correlation scores of all associated combinations, and issue early warnings based on the type of early warning fault labels. The SOP triggering and recording closed-loop management module is used to obtain fault handling plan data based on the warning content and execute SOP triggering. All coordinated operations during the fault handling process are recorded in detail to form a complete fault handling plan. The entire fault handling process is tracked and evaluated to achieve closed-loop management.

Images