Tracing type power transformation professional evaluation portrait method, system and device based on multi-source data fusion and storage medium

Abstract

The invention relates to the technical field of power system evaluation and data analysis, in particular to a multi-source data fusion traceability type power transformation professional evaluation portrait method, system and device and a storage medium. The method comprises the following steps: establishing a time-space-index three-dimensional data model by collecting equipment state data, operation and maintenance management data and safety fault data related to a power transformation specialty to realize unified time-space fusion processing of multi-source heterogeneous data; establishing a three-level evaluation index system, and determining a combination weight through combination of an analytic hierarchy process and an entropy weight method; calculating a comprehensive score by adopting an ideal solution sorting method, verifying the ranking stability through grey correlation analysis, and generating hierarchical evaluation rankings of the power supply bureau and the power transformation management office; tracing type problem diagnosis is adopted, a causal inference framework is combined to verify a causal relationship between associated features and evaluation results, and a root node importance score is calculated based on depth-first search. The problems that a traditional power transformation professional evaluation method is single in data source, incomplete in evaluation dimension, lack of a traceability mechanism and the like are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system evaluation and data analysis, and in particular relates to a multi-source data fusion traceability type substation professional evaluation portrait method, system, device and storage medium. BACKGROUND

[0002] As a key link of power grid safe and stable operation, the substation professional of the power system undertakes the important responsibility of power transformation, distribution and protection. With the continuous expansion of the power grid scale and the sharp increase in the number of equipment, the substation professional management is facing unprecedented challenges. The traditional substation professional evaluation method mainly relies on manual experience and single data source analysis, which has the problems of single evaluation dimension, strong subjectivity, poor timeliness, etc., and is difficult to adapt to the complexity and dynamics of modern power grid.

[0003] In the current substation professional management, the data generated by various systems presents the characteristics of multi-source heterogeneity, including equipment monitoring data, operation and maintenance record data, safety management data, etc. These data are scattered in different information systems, forming a serious "data island" phenomenon. Due to the lack of effective data fusion technology, the internal correlation between the data cannot be fully mined, resulting in the lack of comprehensiveness and accuracy of the evaluation results. At the same time, the traditional evaluation method mainly uses static index system and fixed weight distribution, which cannot adapt to the dynamic changes of the power grid operation environment.

[0004] The existing substation professional evaluation technology also has the problem of insufficient traceability. When the evaluation result shows that there is a problem in a power supply bureau or substation, it often only stays at the phenomenon level description, and it is difficult to deeply mine the root cause of the problem. This shallow analysis method not only affects the efficiency of problem solving, but also may lead to the situation of treating the symptoms but not the root cause, and cannot fundamentally improve the management level of the substation professional. SUMMARY

[0005] In view of the problems existing in the prior art, the present application is proposed.

[0006] Therefore, the main problem solved by the present application is that the existing substation professional evaluation method has the problems of single data source, incomplete evaluation dimension, lack of effective traceability mechanism, etc., which leads to low accuracy of the evaluation result, inaccurate problem positioning, and insufficient decision support capability. Specific problems include: multi-source heterogeneous data cannot be effectively fused and utilized, static evaluation index system has poor adaptability, evaluation results lack objective verification, problem traceability stays at the correlation analysis level and cannot realize deep causal inference, etc.

[0007] To solve the above technical problems, the present application provides the following technical solutions:

[0008] In a first aspect, the embodiments of the present application provide a multi-source data fusion traceability type substation professional evaluation portrait method, which comprises collecting multi-source data related to substation professionals, establishing a time-space-index three-dimensional data model, and performing unified space-time fusion processing on the multi-source data;

[0009] A three-level evaluation index system is established, a target layer-criterion layer-index layer three-level architecture is adopted, and the weight is determined by combining subjective weight calculation and objective weight calculation;

[0010] An ideal solution sorting method is used to calculate the comprehensive score, determine the positive ideal solution and the negative ideal solution, calculate the distance of each evaluation object to the ideal solution and generate the comprehensive score, verify the ranking stability through gray correlation analysis, and generate the hierarchical evaluation ranking of the power supply bureau and the substation management office;

[0011] A traceability type problem diagnosis is adopted, the causal relationship between the associated features and the evaluation results is verified by combining a causal inference framework, and finally the importance ranking of the root nodes is output, and the importance score of the root nodes is calculated based on the path length and the relationship strength.

[0012] As a preferred scheme of the multi-source data fusion traceability type substation professional evaluation portrait method, the collected multi-source data includes equipment state data, operation and maintenance management data, and safety fault data; the equipment state data includes transformer insulation resistance, the operation and maintenance management data includes inspection completion rate, defect processing time length and maintenance quality qualification rate, and the safety fault data includes the number of illegal operation, fault tripping rate and accident processing time length;

[0013] The collected multi-source data is subjected to unified space-time fusion processing, the data of different sampling frequencies is unified to a minute time slice, and a unique space identifier containing provincial and municipal substation equipment information is established for each data entry, so that data quality evaluation is obtained.

[0014] The beneficial effects of the preferred technical scheme are that by specifying the specific types of multi-source data and unified space-time fusion processing, the accuracy of the evaluation basic data is ensured in the early data collection stage, and the problem of single data source in the traditional evaluation method is solved. The standardization processing of heterogeneous data is realized by the minute time slice alignment and the unique space identifier establishment in the middle stage space-time fusion processing. The data quality evaluation mechanism in the later stage ensures that only high-quality data enters the evaluation process through the comprehensive score of integrity, accuracy and timeliness.

[0015] As a preferred scheme of the multi-source data fusion-based traceable power transformation professional evaluation portrait method, the three-level evaluation index system establishment comprises: constructing a judgment matrix and calculating a subjective weight by an analytic hierarchy process, calculating an objective weight based on data distribution characteristics by an entropy weight method, and combining the subjective weight and the objective weight to generate a final combined weight by a multiplication synthesis method; the target layer is a power transformation professional comprehensive evaluation score, the criterion layer includes equipment health degree, operation and maintenance efficiency, safety risk, and economic efficiency, and the index layer is a classification basic index under the criterion layer.

[0016] As a preferred scheme of the multi-source data fusion-based traceable power transformation professional evaluation portrait method, the three-level evaluation index system establishment comprises: constructing a judgment matrix and calculating a subjective weight by an analytic hierarchy process, calculating an objective weight based on data distribution characteristics by an entropy weight method, and combining the subjective weight and the objective weight to generate a final combined weight by a multiplication synthesis method; the target layer is a power transformation professional comprehensive evaluation score, the criterion layer includes equipment health degree, operation and maintenance efficiency, safety risk, and economic efficiency, and the index layer is a classification basic index under the criterion layer.

[0017] As a preferred scheme of the multi-source data fusion-based traceable power transformation professional evaluation portrait method, the three-level evaluation index system establishment comprises: constructing a judgment matrix and calculating a subjective weight by an analytic hierarchy process, calculating an objective weight based on data distribution characteristics by an entropy weight method, and combining the subjective weight and the objective weight to generate a final combined weight by a multiplication synthesis method; the target layer is a power transformation professional comprehensive evaluation score, the criterion layer includes equipment health degree, operation and maintenance efficiency, safety risk, and economic efficiency, and the index layer is a classification basic index under the criterion layer.

[0018] The beneficial effects of the preferred technical scheme are: by establishing the traceable problem diagnosis, the threshold screening is used to accurately identify the to-be-diagnosed object in the early stage, so as to avoid invalid analysis on the normally running unit. In the middle stage, the association rule algorithm is used to mine the association relationship between the low-score index and the bottom layer data, and the feature importance analysis is introduced to screen the key influence features, so as to quickly locate the key problem factors from the massive data. In the later stage, the causal relationship is verified by combining the causal inference framework, so as to avoid the false correlation problem that may be caused by the traditional association analysis, and to ensure the accuracy of problem tracing.

[0019] As a preferred scheme of the multi-source data fusion-based traceable power transformation professional evaluation portrait method, the three-level evaluation index system establishment comprises: constructing a judgment matrix and calculating a subjective weight by an analytic hierarchy process, calculating an objective weight based on data distribution characteristics by an entropy weight method, and combining the subjective weight and the objective weight to generate a final combined weight by a multiplication synthesis method; the target layer is a power transformation professional comprehensive evaluation score, the criterion layer includes equipment health degree, operation and maintenance efficiency, safety risk, and economic efficiency, and the index layer is a classification basic index under the criterion layer.

[0020] As a preferred scheme of the multi-source data fusion-based traceable power transformation professional evaluation portrait method, the three-level evaluation index system establishment comprises: constructing a judgment matrix and calculating a subjective weight by an analytic hierarchy process, calculating an objective weight based on data distribution characteristics by an entropy weight method, and combining the subjective weight and the objective weight to generate a final combined weight by a multiplication synthesis method; the target layer is a power transformation professional comprehensive evaluation score, the criterion layer includes equipment health degree, operation and maintenance efficiency, safety risk, and economic efficiency, and the index layer is a classification basic index under the criterion layer.

[0021] In a second aspect, the embodiment of the present application provides a multi-source data fusion traceability type substation professional evaluation portrait system, which comprises a data fusion module, a multi-source data related to substation professional is collected, a time-space-index three-dimensional data model is established, and unified space-time fusion processing is performed on the multi-source data;

[0022] An evaluation system module establishes a three-level evaluation index system, adopts a target layer-criterion layer-index layer three-level architecture, and determines the weight by combining subjective weight calculation and objective weight calculation;

[0023] An evaluation portrait module adopts an ideal solution sequencing method to calculate a comprehensive score, determines a positive ideal solution and a negative ideal solution, calculates the distance of each evaluation object to the ideal solution and generates a comprehensive score, verifies the ranking stability through grey correlation analysis, and generates a hierarchical evaluation ranking of the power supply bureau and the substation management office.

[0024] A traceability diagnosis module adopts traceability type problem diagnosis, verifies the causal relationship between the correlation characteristics and the evaluation results in combination with a causal inference framework, finally outputs the importance ranking of the root nodes, and calculates the importance score of the root nodes based on the path length and the relationship strength.

[0025] In a third aspect, the embodiment of the present application provides a computer device comprising a memory and a processor, and the memory stores a computer program, wherein the computer program instructions are executed by the processor to realize the steps of the multi-source data fusion traceability type substation professional evaluation portrait method according to the first aspect of the present application.

[0026] In a fourth aspect, the embodiment of the present application provides a computer readable storage medium, which stores a computer program, wherein the computer program instructions are executed by the processor to realize the steps of the multi-source data fusion traceability type substation professional evaluation portrait method according to the first aspect of the present application.

[0027] The present application has the following beneficial effects: by constructing a time-space-index three-dimensional data model, the present application solves the problem that multi-source heterogeneous data cannot be uniformly fused in substation professional management, by establishing a dynamically configurable three-level evaluation index system, adopting a combined weight determination method of subjective weight and objective weight, and avoiding the problem that static weight distribution cannot adapt to the dynamic changes of power grid operation environment. Through double model driven evaluation, the ideal solution sequencing method is used to calculate the comprehensive score, the grey correlation analysis is used to verify the ranking stability, the objectivity and reliability of the evaluation results are ensured, through traceability type problem diagnosis, the deep traceability from correlation analysis to causal inference is realized, the problem root is accurately located, and the deficiency that the traditional evaluation method can only stay on the phenomenon level is solved. BRIEF DESCRIPTION OF DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0029] Figure 1 Flow chart of the traceable substation professional evaluation portrait method for multi-source data fusion;

[0030] Figure 2 Computer device diagram of the traceable substation professional evaluation portrait method for multi-source data fusion;

[0031] Figure 3 Traceable evaluation portrait implementation path schematic diagram of the traceable substation professional evaluation portrait method for multi-source data fusion;

[0032] Figure 4 Data analysis and processing flow logic diagram of the traceable evaluation portrait of the traceable substation professional evaluation portrait method for multi-source data fusion;

[0033] Figure 5 Data input layer schematic diagram of the traceable substation professional evaluation portrait method for multi-source data fusion;

[0034] Figure 6 Evaluation portrait clustering analysis diagram of the traceable substation professional evaluation portrait method for multi-source data fusion;

[0035] Figure 7 Guizhou power grid substation professional traceable evaluation portrait data architecture schematic diagram of the traceable substation professional evaluation portrait method for multi-source data fusion;

[0036] Figure 8 Traceable evaluation portrait cloud view board visual display interface schematic diagram of the traceable substation professional evaluation portrait method for multi-source data fusion;

[0037] Figure 9 Traceable evaluation portrait cloud view board visual display interface schematic diagram of the traceable substation professional evaluation portrait method for multi-source data fusion. DETAILED DESCRIPTION

[0038] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification.

[0039] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details set forth in this description. In other instances, well-known methods, procedures, components, and circuits have not been described in detail so as not to unnecessarily obscure aspects of the present application.

[0040] Secondly, the "one embodiment" or "an embodiment" referred to herein means a specific feature, structure, or characteristic under discussion. The appearance of "in one embodiment" at various places in the specification does not necessarily refer to the same embodiment, nor does it necessarily refer to a single alternative embodiment.

[0041] Embodiment 1

[0042] Reference Figure 1 - Figure 2 For the first embodiment of the present application, the embodiment provides a multi-source data fusion traceability type substation professional evaluation portrait method, comprising,

[0043] S100: Collecting substation professional related multi-source data, and establishing a time-space-index three-dimensional data model to perform unified space-time fusion processing on the multi-source data;

[0044] S200: Establishing a three-level evaluation index system, adopting a target layer-criterion layer-index layer three-level architecture, and determining the weight by combining subjective weight calculation and objective weight calculation;

[0045] S300: Calculating the comprehensive score by using the ideal solution sorting method, determining the positive ideal solution and the negative ideal solution, calculating the distance of each evaluation object to the ideal solution and generating the comprehensive score, verifying the ranking stability by gray correlation analysis, and generating the hierarchical evaluation ranking of the power supply bureau and the substation management office;

[0046] S400: Adopting traceability type problem diagnosis, verifying the causal relationship between the correlation characteristics and the evaluation results in combination with the causal inference framework, finally outputting the importance ranking of the root nodes, and calculating the importance score of the root nodes based on the path length and the relationship strength.

[0047] It should be noted that there are key technical problems in substation professional management, such as scattered data sources, non-uniform evaluation standards, and difficult problem tracing. The traditional evaluation method relies on a single data source and manual experience judgment, and cannot integrate heterogeneous data from multiple business systems such as SCADA system, production management system PMS, and safety supervision system, resulting in a lack of comprehensiveness and objectivity of the evaluation results. At the same time, the existing evaluation system mostly adopts static weight distribution, which cannot adapt to the dynamic changes of the power grid operation environment, and after problem identification, it can only stay at the phenomenon level analysis, lacking the depth of causal tracing ability, and it is difficult to provide accurate root positioning and improvement guidance for management decision-making.

[0048] Therefore, in view of the above-mentioned problems of single evaluation dimension, insufficient traceability, and inaccurate decision support, the present application constructs a time-space-index three-dimensional data model through S100 to realize unified fusion processing of multi-source heterogeneous data, establishes a dynamically configurable three-level evaluation index system through S200, improves the accuracy of the evaluation results by using a subjective and objective weight combination method, adopts a double-model-driven evaluation ranking through S300, verifies the stability of the results by using gray correlation analysis, realizes deep traceability from correlation analysis to causal inference through S400, and accurately locates the root cause of the problem.

[0049] The main technical terms involved in the technical solutions of the present application are explained as follows:

[0050] SCADA: Supervisory-Control-and-Data-Acquisition, data acquisition and monitoring control system; PMS: Production-Management-System, production management information system; TOPSIS: Technique-for-Order-Preference-by-Similarity-to-an-Ideal-Solution, ideal solution ordering method; AHP: Analytic-Hierarchy-Process, analytic hierarchy process; PSM: Propensity-Score-Matching, propensity score matching; DFS: Depth-First-Search, depth-first search algorithm.

[0051] Example 2

[0052] Referring to Figure 3 - Figure 9 , which is a second embodiment of the present application.

[0053] In the present embodiment, the collected multi-source data is subjected to unified spatio-temporal fusion processing in step S100, including the following steps A1-A2:

[0054] A1: The collected multi-source data includes equipment state data, operation and maintenance management data, and safety fault data; the equipment state data includes transformer insulation resistance, the operation and maintenance management data includes inspection completion rate, defect processing time length, and maintenance quality qualification rate, and the safety fault data includes the number of illegal operation, fault tripping rate, and accident processing time length;

[0055] Specifically, the collection of equipment state data is realized by real-time data acquisition through the power grid SCADA system, substation automation system and online detection device. The transformer insulation resistance data is automatically collected by the insulation monitoring device every hour, and the collected parameters include winding-to-ground insulation resistance, winding-to-winding insulation resistance and oil dielectric loss value. The data format is a structured record of timestamp, equipment number, measurement type, value and unit. The circuit breaker mechanical property data is obtained through the operating mechanism monitoring system, including key parameters such as opening and closing time, speed curve, travel characteristics and operation frequency, and the data integrity is ensured by automatic recording after each operation. The GIS equipment SF6 gas pressure is continuously monitored by the density relay and online monitoring system, with a monitoring frequency of once per minute, recording pressure value, temperature compensation value and alarm state information.

[0056] The collection of operation and maintenance management data relies on the production management information system PMS and mobile operation terminal to realize full-process recording. The inspection completion rate data is automatically counted through the GPS positioning and two-dimensional code scanning functions of the mobile inspection terminal. The system records the inspection personnel, inspection time, inspection path and completion status, and compares them with the standard inspection plan to calculate the completion rate. The defect processing time data is timed from the defect discovery input system, recording the defect level, processing personnel, processing measures and completion time. The system automatically calculates the total time consumed from discovery to elimination. The maintenance quality qualification rate data is obtained by statistical analysis of the quality inspection results in the maintenance acceptance process, including maintenance project completion, quality inspection score and acceptance conclusion. The system monthly calculates the proportion of qualified projects to total maintenance projects.

[0057] The collection of safety fault data is realized by combining the safety supervision system and the accident management system to achieve multi-source convergence. The number of illegal operations is identified and recorded through on-site inspection by safety supervision personnel, video monitoring analysis and operation ticket review. Each illegal operation is marked with the type of illegal operation, severity, responsible personnel and rectification measures. The fault tripping rate data is automatically extracted from the dispatching automation system and protection devices, recording the tripping time, fault type, impact range and recovery time. The system classifies and counts the tripping frequency according to the device type and voltage level. The accident processing time data covers the whole process from fault occurrence to complete power restoration, including key nodes such as fault discovery time, processing response time, on-site disposal time and recovery power transmission time, which are recorded in detail through dispatching logs and on-site reports.

[0058] A2: Perform unified spatio-temporal fusion processing on the collected multi-source data, unify the data of different sampling frequencies to minute time slices, and establish a unique spatial identifier containing provincial and municipal substation equipment information for each data entry to obtain data quality evaluation.

[0059] Specifically, all data is uniformly sampled to "minute time slices" in combination with the needs of the substation professional business. For sampling intervals of 1 minute, 5 minutes, 15 minutes and 60 minutes, the data is uniformly sampled to 1 minute time slices. High-frequency data After downsampling The numerical formula for each time slice is as follows:

[0060]

[0061] In a data warehouse, spatial identifiers are used to establish relationships between different data items, and a unique spatial identifier is added to each data entry (province / city substation equipment). This enables spatial tracing.

[0062] The obtained multi-dimensional data is preprocessed to ensure data reliability.

[0063] The detection was performed using a combination of "Isolation Forest" and "DBSCAN Clustering" algorithms.

[0064] For single-point anomaly detection, the Isolation Forest method is used; for cluster anomaly detection, the DBSCAN clustering method is used. The core formula is: for sample... If its - Number of samples within the neighborhood ≥ number of minimum core points ,but The core point; if the sample is at the core point - If it is within the neighborhood, it is a boundary point; otherwise, it is an outlier (removed).

[0065] After detecting abnormal data, different methods are used to fill in missing values ​​according to the data type.

[0066] For time-series data, linear interpolation is used for filling, and the formula is: If ,and If not missing, then For non-time-series data, KNN interpolation is used to fill the gaps.

[0067] The data is normalized using Z-score standardization. After normalization, spatiotemporal consistency verification is performed.

[0068] Check the difference between the data timestamp and the collection time. If the difference is greater than 30 seconds, mark it as "time abnormal" and refill the data.

[0069] Verify spatial identifiers Check if the device exists in the power grid equipment ledger. If it does not exist, mark it as "invalid space" and refuse to add it to the database.

[0070] The processed data was scored based on "completeness (40%) + accuracy (30%) + timeliness (30%)". ,only The data will be used for subsequent evaluation.

[0071] The three categories of indicators are as follows:

[0072] Integrity (40%): Number of missing data / Total number of data (For example, 100 data missing 5 data, integrity );

[0073] Accuracy (30%): Number of abnormal data / Total number of data (For example, 100 data containing 3 abnormal data, accuracy );

[0074] Timeliness (30%): Number of overdue data / Total number of data (For example, 100 data containing 2 overdue data, timeliness );

[0075] In an optional embodiment, the unified spatio-temporal fusion processing of the collected multi-source data in step S100 further includes realizing parallel acquisition of multi-system data through a distributed data acquisition architecture. A unified data exchange interface is established, and each business system pushes data to the fusion platform in real time through a standardized data bus. The data bus adopts a publish-subscribe mode to ensure the real-time and reliability of data transmission. For different voltage level of power transformation equipment, differentiated data acquisition frequency is set according to the importance of the equipment. 220kV and above equipment adopts minute-level acquisition, 110kV equipment adopts 5-minute-level acquisition, and 35kV and below equipment adopts 15-minute-level acquisition, which reduces the system load while ensuring the timeliness of data.

[0076] In another optional embodiment, the unified spatio-temporal fusion processing of the collected multi-source data in step S100 further includes establishing data bloodline tracing. Metadata information such as data source identifier, acquisition time, processing time and data version number is added to each data record to form a complete data life cycle tracing chain. When data quality assessment finds abnormalities, the source system and acquisition node of the abnormal data are quickly located through bloodline tracing, and the source data reacquisition process is triggered. At the same time, a data quality monitoring large screen is established to display the data integrity, accuracy and timeliness indicators of each data source in real time, and automatically send an alarm notification when any indicator is below the preset threshold.

[0077] In the present embodiment, the three-level evaluation index system is established in step S200, including the following steps B1-B2:

[0078] B1: Constructing a judgment matrix and calculating subjective weights by AHP method, calculating objective weights based on data distribution characteristics by entropy method, and combining subjective weights and objective weights to generate final combined weights by multiplication synthesis method;

[0079] B2: The target layer is the substation comprehensive evaluation score, the criterion layer includes equipment health, operation and maintenance efficiency, safety risk and economic efficiency, and the index layer is the classification basic index under the criterion layer.

[0080] Specifically, the criterion layer includes 4 indexes, equipment health ( ), operation and maintenance efficiency ( ), safety risk ( ), and economic efficiency ( ); the index layer is the classification basic index under the criterion layer.

[0081] Subjective weight calculation is performed by AHP.

[0082] Construct a judgment matrix , , , , ,

[0083] , , , , , ,

[0084] , , , ,

[0085] , ,

[0086] , , , , ,

[0087] , , , , , , ,

[0088] , , , ,

[0089] , , , ,

[0090] In an alternative embodiment, the establishment of the three-level evaluation index system in step S200 further includes establishing dynamic adjustment of the index system. The system periodically analyzes the discrimination and sensitivity of each index. For indexes with a discrimination lower than 0.3, mark them as candidate elimination indexes. For indexes with too high sensitivity leading to large fluctuations in evaluation results, reduce the weight. Through regression analysis of historical evaluation data, identify potential factors that affect the comprehensive score but are not included in the index system. After expert review, include them in the index system to form dynamic updates of the index library. At the same time, establish self-adaptive adjustment of the index early warning threshold. According to the data distribution characteristics of the past three months, automatically update the normal interval and abnormal threshold of each index.

[0091] In another alternative embodiment, the establishment of the three-level evaluation index system in step S200 further includes optimizing the expert weight determination process using the Delphi method. Organize experts in power transformation, operation and maintenance management personnel, and technical personnel to form an expert group. Through multiple rounds of anonymous questionnaire surveys, collect experts' judgments on the relative importance of each index. Calculate the dispersion of expert opinions in each round of questionnaires. When the dispersion converges to a preset threshold, terminate the questionnaire survey. Combine the opinions of experts in each round to form the final subjective weight.

[0092] In this embodiment, the ideal solution ranking method is used to calculate the comprehensive score in step S300, including the following step C1:

[0093] C1: Verify the ranking stability by gray correlation analysis. The gray correlation analysis takes the positive ideal solution as the reference sequence to calculate the correlation coefficient and correlation degree. When the correlation degree reaches the preset threshold, confirm the effectiveness of the evaluation results.

[0094] Specifically, a double-model-driven professional portrait evaluation and ranking is adopted, which combines "TOPSIS comprehensive score" and "gray correlation ranking verification".

[0095] Map the fused data to the index system and calculate the score according to the index type:

[0096] Positive index: ( Normalized value).

[0097] Negative index: (Reverse conversion and weighting).

[0098] TOPSIS comprehensive score calculation.

[0099] Construct a weighted standardized matrix , ;

[0100] Determine the dynamic ideal solution:

[0101] Positive ideal solution ;

[0102] Negative ideal solution ;

[0103] Calculate the distance and comprehensive score, and map the results to 0-100 points.

[0104] Distance , ;

[0105] Comprehensive score .

[0106] Verify the stability of the TOPSIS ranking through gray correlation analysis. The steps are as follows:

[0107] Take the positive ideal solution as the reference sequence ;

[0108] Calculate the correlation coefficient :

[0109]

[0110] Where is the resolution coefficient (take 0.5);

[0111] Calculate the correlation degree ;

[0112] Set the Spearman correlation coefficient threshold to 0.85, which is verified based on historical data: when the correlation coefficient is ≥0.85, the difference in ranking between the two methods has less than 5% impact on business decisions, and can be considered as effective ranking; otherwise, trigger the weight recalibration process to ensure the reliability of the evaluation results.

[0113] Based on the criterion layer scores ( ), ( ), ( ), ( ), use ECharts to generate radar charts and superimpose the criterion layer score trend curves in the past 3 months to visually display the strengths and weaknesses of the evaluation objects.

[0114] In an alternative embodiment, the ideal solution ranking method is used to calculate the comprehensive score in step S300, and further includes introducing time series analysis to identify the trend characteristics of the evaluation objects. The score change rate and fluctuation coefficient of each evaluation object are calculated for the past six months. The objects with a continuously rising score are marked as "progressive", the objects with a continuously falling score are marked as "regressive", and the objects with a large score fluctuation are marked as "unstable". In the final ranking, the "progressive" objects are given trend bonus, and the "regressive" objects are given trend penalty, encouraging continuous improvement and warning of performance decline. At the same time, the time series prediction curve of the evaluation objects is generated, predicting the score trend in the next three months to provide decision basis for early intervention.

[0115] In another alternative embodiment, the ideal solution ranking method is used to calculate the comprehensive score in step S300, and further includes establishing a hierarchical evaluation. All evaluation objects are divided into different levels according to the equipment scale, power supply load and management complexity. The objects within the same level are compared and ranked horizontally, and the objects between different levels are not directly compared. For each evaluation object within a level, a differentiated ideal solution and negative ideal solution are set to ensure that the evaluation standard matches the actual management difficulty. Through level coefficient adjustment, low-level objects are allowed to be promoted to high-level after reaching a certain standard in comprehensive score, encouraging evaluation objects to continuously improve management level.

[0116] In the present embodiment, the traceability problem diagnosis in step S400 includes the following steps D1-D3:

[0117] D1: Mark the evaluation objects with a comprehensive score or single criterion layer score below a threshold value as objects to be diagnosed, and use the association rule algorithm to mine the association relationship between the low-score indicators and the underlying data, and introduce feature importance analysis to screen the key influencing features.

[0118] Specifically, the traceability problem diagnosis is achieved by multi-dimensional fusion of "association rule + causal inference".

[0119] Mark the evaluation objects with a "comprehensive score " or "single criterion layer score " as "objects to be diagnosed", and extract the low-score indicator set threshold value .

[0120] The Apriori algorithm is used to mine the association relationship between the low-score indicators and the underlying data, and the XGBoost feature importance is introduced to screen the key influencing features. An XGBoost classification model is constructed: taking "whether it is a low-score indicator" as the label (1 = yes, 0 = no), and the underlying data as the features; output the feature importance score , and screen the features as "highly associated features".

[0121] Further, the identification of the to-be-diagnosed object adopts multi-level threshold screening, and the system sets an evaluation object with a comprehensive score lower than 70 points or a score of any criterion layer lower than 60 points as the to-be-diagnosed object. The threshold setting is based on statistical analysis of historical data, and the comprehensive score of 70 points corresponds to the 25th percentile of the scores of all evaluation objects, and the single criterion layer score of 60 points corresponds to the 20th percentile of the scores of each criterion layer. In the identification process, the system records the specific values, deviation degrees and historical change trends of the low-score indicators to form a problem indicator list.

[0122] The association rule mining adopts an improved Apriori algorithm, and sets the minimum support threshold to 0.1 and the minimum confidence threshold to 0.6. The algorithm first scans all the underlying data of the to-be-diagnosed objects, identifies frequently occurring data patterns, and then calculates the association strength between each data indicator and the low-score evaluation result. The representation form of the association rule is “IF data condition THEN low-score result”, for example, “IF transformer oil temperature > 85°C AND load rate > 90% THEN equipment health degree score < 60”. The system outputs all the association rules that meet the threshold conditions and sorts them according to the confidence and lift.

[0123] After the training is completed, the system extracts the importance score of each feature, and the importance score reflects the contribution degree of the feature to the prediction result. The features with a feature importance score greater than 0.1 are defined as high-correlation features, and these features will enter the subsequent causal inference analysis link. The feature importance analysis also outputs the correlation matrix between the features to identify the feature combinations with multicollinearity, which provides a basis for the identification of confounding variables in causal inference.

[0124] D2: further comprising verifying the causal relationship between the high-correlation features and the low-score indicators based on the causal inference framework, determining the causal effect strength of the influencing factors through confounding variable control and effect calculation.

[0125] Specifically, the causal relationship between the high-correlation features and the low-score indicators is verified based on the DoWhy framework, and the steps are as follows:

[0126] define the causal hypothesis, → ; define the confounding variables through power grid domain knowledge , draw the causal graph ( as the edges between variables);

[0127] adopt “propensity score matching (PSM)” to control the confounding variables, calculate the propensity score (based on Logistic regression, with as the independent variable, as the dependent variable), and match 1 for each And it tends to favor samples with similar scores; calculate the average treatment effect (ATE):

[0128]

[0129] Combination ,determination for The causal factors.

[0130] Furthermore, the causal inference framework employs the four-step methodology of the DoWhy library to achieve deep causal analysis. The first step is causal hypothesis modeling. The system constructs a causal hypothesis graph based on power grid expertise, identifying processing variables (highly correlated features), outcome variables (low-scoring indicators), and confounding variables (third-party factors that may affect both). The identification of confounding variables is based on the power grid operating mechanism. For example, when analyzing the impact of transformer temperature on equipment health, ambient temperature, load level, and cooling system status may all be confounding variables.

[0131] The second step is the construction of the causal graph. The system uses a directed acyclic graph (DAG) to represent the causal relationships between variables, with each edge in the graph representing a hypothesis of a causal relationship. The construction of the causal graph combines the operating principles of power grid equipment, historical fault case analysis, and expert knowledge to ensure the rationality and completeness of the graph structure. The system verifies the identifiability of the causal graph to ensure the existence of causal effect estimation paths.

[0132] The third step is causal effect identification. The system uses methods such as the backdoor criterion and the frontdoor criterion to identify estimable causal effects. For cases that meet the backdoor criterion, the system selects the minimum sufficient set of confounding variables for control; for cases that do not meet the backdoor criterion, the system attempts to find mediating variables to achieve frontdoor criterion estimation. The identification process outputs the required variable control strategies and estimation method suggestions.

[0133] The fourth step is causal effect estimation. The system uses the propensity score matching (PSM) method to control for the influence of confounding variables. First, a logistic regression model is used to calculate the propensity score for each observation unit to receive treatment. The model includes all identified confounding variables. Then, a 1:1 nearest neighbor matching method is used to match each treatment unit with a control group unit whose propensity score is closest to the treatment unit's, with a matching tolerance of 0.01. After matching, the system calculates the average treatment effect (ATE) using the following formula:

[0134]

[0135] Where Y is the outcome variable and X is the treatment variable. The causal relationship is verified by a paired t-test, with a p-value less than 0.05 considered causality. The system also calculates 95% confidence intervals to provide an assessment of the uncertainty of the causal effect.

[0136] D3: further comprising, traversing the associated path of the causal factors and the problem by using depth-first search, calculating the importance score of the root node based on the path length and the relationship strength, and outputting the root cause analysis result of the core root cause factor.

[0137] Specifically, the associated path of the causal factors and the problem is traversed by using depth-first search (DFS). The associated path of the causal factors-low index is traversed by using depth-first search (DFS), and the maximum traversal depth is 3; the importance of the root node is calculated based on the path length and the relationship strength:

[0138]

[0139] Output the node as the core root.

[0140] Further, the implementation of the depth-first search algorithm uses recursive traversal, starting from each verified causal factor as a starting node. During the search process, the system maintains an access state table to record the access state (not accessed, accessing, completed) of each node, avoiding loop traversal. The search depth is limited to 3 levels, i.e., the maximum of 3 levels of influence factors are traced back from the starting causal factor. The recording format of each path is [starting node, intermediate node 1, intermediate node 2, terminal node], and the system records the relationship strength value between each pair of adjacent nodes in the path.

[0141] The calculation of the relationship strength is based on the comprehensive score of the causal inference result and the association rule strength. For the relationship verified by causality, the relationship strength is equal to the absolute value of the causal effect; for the relationship only analyzed by association, the relationship strength is equal to the confidence of the association rule x the lift. The path strength is calculated by using the geometric mean method to avoid the excessive influence of a single weak relationship on the overall path, and the calculation formula is path strength=(∏ relationship strengthi)^(1 / n), where n is the number of relationships in the path.

[0142] The calculation of the importance score of the root node considers how many paths the node appears in, the strength distribution of these paths, and the influence of path length. The calculation formula is:

[0143]

[0144] The weight of the path length reflects the principle that "the closer the distance, the greater the influence". The system also calculates the influence breadth of each root node, i.e., the number of downstream indicators that the node can affect, and the node with high influence breadth is considered to be a more critical root cause factor.

[0145] The output traceability analysis result includes core root cause ranking, specific causal transmission path diagram and targeted improvement suggestions. The core root causes are ranked in descending order of importance score, and nodes with a score greater than 0.7 are marked as root causes that need to be focused on. The causal transmission path diagram is displayed in the form of a directed graph, with node size reflecting importance score, edge thickness reflecting relationship strength, and color depth indicating path level. Improvement suggestions are automatically generated based on the type and impact mechanism of the root cause, including specific control measures, responsible departments and expected effects, etc., providing actionable guidance for management decisions.

[0146] In an optional implementation, the traceability problem diagnosis in step S400 further includes establishing a problem early warning preposition. Risk assessment is performed on all evaluation objects, and the comprehensive score decline rate and criterion layer score abnormal fluctuation rate are calculated. When the decline rate exceeds the threshold or the abnormal fluctuation rate exceeds the threshold for three consecutive months, the object is included in the early warning monitoring range. Through machine learning model prediction of the future score trend of the evaluation object, the problem diagnosis process is started in advance for the object whose predicted score may fall below the threshold, realizing the transition from post-event analysis to pre-event prevention. At the same time, problem rectification tracking is established, and rectification plans are developed for the identified root causes and the rectification progress is tracked to verify the rectification effect.

[0147] In another optional implementation, the traceability problem diagnosis in step S400 further includes introducing knowledge graph technology to construct a substation professional problem traceability knowledge base. The causal relationships, root causes and improvement measures identified in historical traceability analysis are stored in the form of a knowledge graph, forming a complete knowledge chain of "problem phenomenon - impact factor - root cause - solution". When a new object to be diagnosed appears, similar problem patterns are first searched in the knowledge graph, and the analysis path and solution verified in history are preferentially adopted. Through continuous learning and updating of the knowledge graph, experience knowledge of substation professional management is accumulated.

[0148] In summary, in the data collection link, through the collaborative work of SCADA systems, substation automation systems and online monitoring devices, through the regular collection of transformer insulation resistance by insulation monitoring devices and the automatic recording of circuit breaker mechanical characteristics by operating mechanism monitoring systems, errors and omissions in manual collection are avoided.

[0149] In terms of data preprocessing, a dual-algorithm joint detection of isolation forest and DBSCAN clustering is adopted to improve the accuracy of abnormal data identification and avoid the missed detection and false detection problems that may exist in a single algorithm. Linear interpolation and KNN interpolation methods are used to fill in missing values of time series and non-time series data respectively, ensuring data integrity while maintaining the time sequence characteristics of the data.

[0150] In terms of weight determination, the analytic hierarchy process is combined with the entropy weight method, which fully utilizes the subjective judgment of expert experience, reflects the objective distribution characteristics of data itself, and avoids the deviation that may be caused by relying solely on subjective weight or objective weight. The final combined weight is calculated by multiplication synthesis method.

[0151] In terms of evaluation result verification, the ranking results of TOPSIS are verified for stability by gray correlation analysis, and the threshold of Spearman correlation coefficient is set to 0.85. When the correlation coefficient does not reach the threshold, the weight recalibration process is automatically triggered.

[0152] In terms of traceability diagnosis, the four-step methodology of the DoWhy framework is used to realize strict causal inference verification, avoiding the false correlation problem that may be caused by traditional correlation analysis.

[0153] Embodiment 3

[0154] The above is a schematic scheme of a multi-source data fusion traceability type substation professional evaluation portrait method. It should be noted that the technical scheme of the multi-source data fusion traceability type substation professional evaluation portrait system belongs to the same concept as the technical scheme of the multi-source data fusion traceability type substation professional evaluation portrait method described above. The technical scheme of the multi-source data fusion traceability type substation professional evaluation portrait system in this embodiment is not described in detail, and can be referred to the description of the technical scheme of the multi-source data fusion traceability type substation professional evaluation portrait method.

[0155] The embodiment also provides a multi-source data fusion traceability type substation professional evaluation portrait system, which comprises:

[0156] The data fusion module collects substation professional related multi-source data, and establishes a time-space-index three-dimensional data model to perform unified space-time fusion processing on the multi-source data.

[0157] The evaluation system module establishes a three-level evaluation index system, adopts a target layer-criterion layer-index layer three-level architecture, and determines the weight by combining subjective weight calculation and objective weight calculation.

[0158] The portrait evaluation module calculates the comprehensive score by using the ideal solution sorting method, determines the positive ideal solution and the negative ideal solution, calculates the distance of each evaluation object to the ideal solution and generates the comprehensive score, verifies the ranking stability by gray correlation analysis, and generates the hierarchical evaluation ranking of the power supply bureau and the substation management office.

[0159] The traceability diagnosis module uses traceability problem diagnosis, combines the causal inference framework to verify the causal relationship between the correlation characteristics and the evaluation results, and finally outputs the importance ranking of the root nodes, and calculates the importance score of the root nodes based on the path length and the relationship strength.

[0160] This embodiment also provides an electronic device suitable for the case of traceable substation professional evaluation and profiling based on multi-source data fusion, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the traceable substation professional evaluation and profiling method based on multi-source data fusion as proposed in the above embodiment.

[0161] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the source-based substation professional evaluation and profiling method for realizing multi-source data fusion as proposed in the above embodiment.

[0162] The storage medium proposed in this embodiment belongs to the same inventive concept as the source-based substation professional evaluation and profiling method for realizing multi-source data fusion proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0163] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0164] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A source-based substation professional evaluation and profiling method based on multi-source data fusion, characterized in that: The method comprises collecting multi-source data related to power transformation, establishing a time-space-index three-dimensional data model, and performing unified space-time fusion processing on the multi-source data; A three-level evaluation index system is established, and a target layer-criterion layer-index layer three-level architecture is adopted to determine the weight by combining subjective weight calculation and objective weight calculation; The ideal solution sorting method is used to calculate the comprehensive score, determine the positive ideal solution and negative ideal solution, calculate the distance of each evaluation object to the ideal solution and generate the comprehensive score, verify the ranking stability by gray correlation analysis, and generate the hierarchical evaluation ranking of the power supply bureau and the power transformation management station; The root node importance score is calculated based on the path length and relationship strength, and the root source factor traceability analysis result is output. 2.The multi-source data fusion-based, traceable, substation specialty evaluation image generation method of claim 1, wherein: The collected multi-source data includes device state data, operation and maintenance management data, and safety fault data; the device state data includes transformer insulation resistance, the operation and maintenance management data includes inspection completion rate, defect processing time length and maintenance quality qualification rate, and the safety fault data includes the number of illegal operation, fault tripping rate and accident processing time length; The collected multi-source data is subjected to unified space-time fusion processing, the data of different sampling frequencies are unified to minute time slices, and a unique spatial identifier containing provincial and municipal power transformation station device information is established for each data item to obtain data quality evaluation. 3.The multi-source data fusion-based traceable substation professional evaluation image method of claim 2, characterized in that: The three-level evaluation index system comprises constructing a judgment matrix and calculating a subjective weight by an analytic hierarchy process, calculating an objective weight based on data distribution characteristics by an entropy weight method, and combining the subjective weight and the objective weight to generate a final combined weight by a multiplication synthesis method; The target layer is a power transformation comprehensive evaluation score, the criterion layer includes device health degree, operation and maintenance efficiency, safety risk and economic efficiency, and the index layer is a classification basic index under the criterion layer. 4.The multi-source data fusion-based, traceable, substation specialty evaluation image generation method of claim 3, wherein: The gray correlation analysis takes the positive ideal solution as a reference sequence to calculate the correlation coefficient and correlation degree, and the evaluation result is confirmed to be effective when the correlation degree reaches a preset threshold.

5. The multi-source data fusion-based provenance transformer professional evaluation portrait method of claim 4, wherein: The traceability problem diagnosis comprises marking evaluation objects with a comprehensive score or a single criterion layer score below a threshold as diagnosis objects, using an association rule algorithm to mine the association relationship between low-score indexes and bottom layer data, and introducing feature importance analysis to screen key influencing features.

6. The multi-source data fusion-based provenance transformer professional evaluation portrait method of claim 5, wherein: Further comprising verifying the causal relationship between high-correlation features and low-score indexes based on a causal inference framework, determining the causal effect strength of influencing factors through confusion variable control and effect calculation.

7. The multi-source data fusion-based provenance transformer professional evaluation portrait method of claim 6, wherein: Further comprising using a depth-first search to traverse the association path of causal factors and problems, calculating the root node importance score based on the path length and relationship strength, and outputting the traceability analysis result of the core root source factor.

8. A multi-source data fusion traceable substation professional evaluation image system based on the multi-source data fusion traceable substation professional evaluation image method of any one of claims 1-7. Further comprising a data fusion module for collecting multi-source data related to power transformation, establishing a time-space-index three-dimensional data model, and performing unified space-time fusion processing on the multi-source data; An evaluation system module for establishing a three-level evaluation index system, adopting a target layer-criterion layer-index layer three-level architecture, and determining the weight by combining subjective weight calculation and objective weight calculation; The image evaluation module calculates the comprehensive score by using the ideal solution ranking method, determines the positive ideal solution and the negative ideal solution, calculates the distance of each evaluation object to the ideal solution and generates the comprehensive score, verifies the ranking stability by using the grey correlation analysis, and generates the hierarchical evaluation ranking of the power supply bureau and the transformer management station; The traceability diagnosis module marks the evaluation object with a comprehensive score lower than a preset threshold as a to-be-diagnosed object, verifies the causal relationship between the correlation features and the evaluation results in combination with a causal inference framework, finally outputs the importance ranking of the root nodes, and calculates the importance score of the root nodes based on the path length and the relationship strength. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to implement the steps of the multi-source data fusion traceability type transformer professional evaluation image method of any one of claims 1-7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to implement the steps of the multi-source data fusion traceability type transformer professional evaluation image method of any one of claims 1-7.

Images