Power fault diagnosis and prediction system based on big data

Abstract

The invention discloses an electric power fault diagnosis and prediction system based on big data, particularly relates to the field of intelligent diagnosis of an electric power system, and aims to solve the problems of fault response delay and insufficient diagnosis precision. Localized intelligent diagnosis of power faults is realized, and stable operation and rapid recovery of a power system are guaranteed; a priority classification and sorting uploading mechanism is combined to ensure that key data is transmitted preferentially, diagnosis errors caused by delay are reduced, and the analysis stability of a central dispatching center is enhanced; through comprehensive judgment of a global deviation aggregation index and a trust stability harmonic measurement index, accurate classification of local node diagnosis controllability is realized, more accurate key data is guaranteed to be obtained, finally, the fault response capability of a system and the safety of a power system are improved, and powerful technical support is provided for intelligent diagnosis and maintenance.

Description

Technical Field

[0001] The present invention relates to the field of intelligent diagnosis of power systems, and more specifically, to a power fault diagnosis and prediction system based on big data. Background Art

[0002] In modern power systems, monitoring and fault diagnosis of power equipment are crucial, and local nodes and the central dispatching center play a key role in this process. As independent monitoring devices distributed in the power network, local nodes are responsible for real-time acquisition and processing of switch quantity information and fault recording information related to power equipment. Switch quantity information records the state changes of equipment, such as the opening and closing of circuit breakers and the action of relays, usually expressed in binary form; fault recording information provides detailed electrical quantity change data when the system is abnormal, such as voltage and current waveforms, which are used to analyze the nature and cause of the fault. The local node performs preliminary fault diagnosis on this data, marks key electrical quantities and important waveform data, and then uploads the marked data to the central dispatching center first. As the core analysis platform, the central dispatching center uses network-wide data and complex models to conduct in-depth diagnosis of key data uploaded by local nodes and generate detailed and accurate final diagnosis results.

[0003] However, in the process of uploading fault recording files from local nodes to the central dispatching center, the existing technology often affects the accuracy of diagnosis due to transmission delays. The recording files contain a large amount of waveform data, and transmission delays may cause the central dispatching center to be unable to obtain complete fault information in a timely manner, thereby reducing the accuracy of fault diagnosis and processing efficiency. In an emergency, transmission delays may hinder the central dispatching center from quickly and accurately analyzing the fault, affecting the rapid recovery of the power system. Therefore, the existing technology has shortcomings in dealing with the problem of transmission delays, especially in the lack of optimization methods for the priority transmission and comprehensive analysis of key data. There is an urgent need for a solution that can effectively deal with transmission delays and improve the accuracy and timeliness of fault diagnosis, so as to reduce diagnostic deviations and enhance the response capabilities of the power system.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a power fault diagnosis and prediction system based on big data, which realizes localized intelligent diagnosis of power faults through multi-level data preprocessing, feature extraction, dynamic pattern matching and adaptive update mechanism, improves the diagnostic accuracy and reliability, and ensures the stable operation and rapid recovery of the power system; combines priority classification and sorting upload mechanism, multi-dimensional comprehensive analysis and dynamic adjustment to ensure that key data is transmitted first in complex fault conditions, reduces diagnostic errors caused by delays or losses, and improves the analysis efficiency of the central dispatching center and the stability of the system; through the comprehensive judgment of the global deviation aggregation index and the trust stability harmonic measurement index, it realizes the accurate classification of the controllability of local nodes, further optimizes the diagnosis of controllable nodes, obtains more accurate key data, and finally, with the support of the dynamic adjustment mechanism, comprehensively improves the system's fault response capability and the safety of the power system, and provides strong technical support for intelligent diagnosis and maintenance to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] Data acquisition module: obtains the switch quantity information and fault recording information corresponding to each local node, applies the local node diagnosis model to perform preliminary diagnosis, marks key electrical quantities and waveform data, and transmits the preliminary diagnosis results and the marked electrical quantities and waveform data to the data upload module;

[0008] Data upload module: Based on the preliminary diagnosis results, upload the marked key electrical quantities and waveform data to the central dispatch center and pass them to the data analysis module;

[0009] Data analysis module: receives key electrical quantities and waveform data, uses pre-built models for diagnosis, generates final diagnosis results, and transmits the results to the comparison analysis module and the fault prediction module;

[0010] Comparative analysis module: compares the preliminary diagnosis results with the final diagnosis results, extracts drift information and reliable information, records deviation data, builds a trust stability harmonic measurement index, classifies the controllability of local nodes, and divides them into three controllable categories: high, medium, and low, and passes the classification results to the adjustment optimization module;

[0011] Adjustment and optimization module: calculates the adjustment coefficient according to the trust stability harmonic measurement index, optimizes the preliminary diagnosis results of the medium controllable local nodes, and transmits the optimized diagnosis results to the model optimization module;

[0012] Model optimization module: Apply adjustment coefficients to optimize the local node diagnosis model, deploy the optimized local node diagnosis model to the key electrical quantity and waveform data extraction tasks in the subsequent collected switch quantity information and fault recording information, and send the optimized local node diagnosis model to the data acquisition module;

[0013] Fault prediction module: compare the final diagnosis results with historical data to extract features, conduct real-time monitoring and identification, calculate the fault probability, and generate early warning signals.

[0014] In a preferred embodiment, the operation process of the data acquisition module includes the following contents:

[0015] A1, each local node continuously collects the switch quantity information and fault recording information related to the corresponding power equipment in real time.

[0016] A2, after collecting the switch quantity information and fault recording information, the local node immediately performs multi-level data preprocessing.

[0017] A3, based on the preprocessed data, first extract the features of the preprocessed switch quantity information and fault recording information; use multi-channel data fusion technology to fuse the feature data of the switch quantity information and fault recording information; after the fusion feature vector is generated, use the dynamic pattern matching algorithm to match the features of the switch quantity information and fault recording information with the pre-stored fault pattern library in real time; based on the real-time matching results, the local node preliminarily identifies the fault type and the time period when the fault occurs, thereby obtaining a preliminary diagnosis result.

[0018] A4, after obtaining the preliminary diagnosis results, the local node marks the key electrical quantity information and waveform data related to the fault.

[0019] In a preferred embodiment, the operation process of the data upload module includes the following contents:

[0020] B1, after the initial diagnosis is completed, first perform importance analysis on the marked electrical quantity information and waveform data.

[0021] B2. Next, the correlation between the marked electrical quantity information and the waveform data is evaluated; through association analysis and feature clustering, the electrical quantities and waveform features that are highly correlated with each other are identified.

[0022] B3, according to the fault type identified in the preliminary diagnosis, the marked electrical quantity information and waveform data are matched and processed; the marked electrical quantity information and waveform data are compared with the preset fault type database to determine which data is the most critical in the corresponding type of fault analysis.

[0023] B4, based on the analysis results of the above steps, a priority score is assigned to each group of marked electrical quantity information and waveform data.

[0024] B5, sort the data from high to low according to the priority score results; the electrical quantity information and waveform data ranked first are marked as key electrical quantity information and waveform data, and are scheduled to be uploaded immediately.

[0025] In a preferred embodiment, the operation process of the data analysis module includes the following contents:

[0026] C1, after the central dispatching center receives the key electrical quantity information and waveform data uploaded by the local node, it first performs a data integrity check.

[0027] C2, after passing the verification, it is standardized.

[0028] C3, feature reconstruction and enhancement after normalization.

[0029] C4, the central dispatch center selects the corresponding pre-built model for diagnosis based on the preliminary diagnosis results of the fault type.

[0030] C5, the central dispatch center runs the selected pre-built models in parallel on multiple computing nodes.

[0031] C6, after generating the preliminary diagnosis results, dynamically adjusts the diagnosis results based on the real-time monitoring data and feedback from other nodes in the central dispatch center to determine the final diagnosis results.

[0032] In a preferred embodiment, the operation process of the comparative analysis module includes the following contents:

[0033] Extract and compare the drift information and reliable information between the preliminary diagnosis results and the final diagnosis results, where the drift information includes the global deviation aggregation index and the reliable information includes the trust stability harmonic measurement index; the global deviation aggregation index and the trust stability harmonic measurement index are comprehensively processed and calculated to obtain the trust stability harmonic measurement index; according to the value of the trust stability harmonic measurement index, the controllability of the local node is classified; the classification criteria are as follows:

[0034] When the value of the trust stability harmonic index is less than the classification threshold one, the node is considered to have high controllability; when the value of the trust stability harmonic index is greater than or equal to the classification threshold one and less than the classification threshold two, the node is considered to have medium controllability; when the value of the trust stability harmonic index is greater than or equal to the classification threshold two, the node is considered to have low controllability.

[0035] In a preferred embodiment, the generation logic and detailed steps of the global deviation aggregation index are as follows:

[0036] D1, first, construct a drift feature vector to quantify the difference between the preliminary diagnosis result and the final diagnosis result; set the drift feature vector D, which contains drift information in multiple dimensions, each dimension corresponds to a feature in the diagnosis result; for the i-th feature, calculate its drift value d i : in, represents the i-th eigenvalue of the final diagnosis result, Represents the i-th eigenvalue of the preliminary diagnosis result.

[0037] D2, to evaluate the reliability of each feature drift value, the fuzzy correlation function μ(d i ), which is used to measure the rationality of the drift value relative to the historical data; the fuzzy correlation function is defined as: Among them, λ i is the threshold parameter set for the i-th feature based on historical data; α is the first adjustment parameter, which controls the steepness of the function.

[0038] D3, according to the fuzzy membership, assign weight w to each feature i ; The weight calculation formula is: Among them, γ is the second adjustment parameter, which controls the influence of drift value on weight; S i Score the importance of the i-th feature; S j Score the importance of the jth feature.

[0039] D4, combine the drift value of each feature with its corresponding weight to calculate the global deviation aggregation index R of the overall diagnosis result; the calculation formula is: Among them, ∈ is a small constant used to avoid the denominator being zero; n represents the number of features.

[0040] In a preferred embodiment, the generation logic and detailed steps of the trust stability harmonic measurement index are as follows:

[0041] E1, construct the drift matrix Q, each element q in the matrix x,y Indicates the relative drift of feature x at time point y; the drift value calculation formula is: in, Represents the value of feature x at time point y in the final diagnosis result; Indicates the corresponding value in the preliminary diagnosis result; is the reference value of the corresponding feature.

[0042] E2, based on the drift matrix, calculates the overall drift accumulation function R(y), which is used to calculate the cumulative drift value of all features at time y; the formula is as follows: Among them, φ(q x,y) is the nonlinear conversion function of the drift value, which is in the form of Where ξ is the third adjustment parameter; p is the total number of features.

[0043] E3, calculates the directionality of the overall drift by constructing a drift direction synthetic vector Z, which represents the drift direction within the entire time range; the formula is as follows: Where sgn(R(y)) is the sign function that determines the drift direction (positive or negative), |R(y)| is the absolute value of the accumulated drift, and T is the total time of fault evolution.

[0044] E4, combining the value of the overall drift accumulation function and the drift direction synthetic vector Z, calculates the final trust stability harmonic measurement index V; the formula is as follows: in, is the fourth adjustment parameter, controlling the impact of cumulative drift on the trust stability harmonic measurement index; The integral of the accumulated drift is used to evaluate the amount of drift over the entire time period.

[0045] In a preferred embodiment, the operation process of the adjustment optimization module includes the following contents:

[0046] F1, first, the trust stability harmonic measurement index is calculated for the preliminary diagnosis results of the local node.

[0047] F2, according to the value of the trust stability harmonic measurement index, preliminarily calculate the adjustment coefficient κ; the calculation formula of the adjustment coefficient is designed as: in, is the fifth adjustment parameter, which is used to control the sensitivity of the adjustment coefficient to the trust stability harmonic measurement index, and ω is a nonlinear adjustment parameter.

[0048] F3, for the preliminary diagnosis results of the medium controllable nodes, the adjustment coefficient is used to optimize them; the optimization process adjusts the key parameters in the diagnosis model to make the final diagnosis results closer to the actual fault situation; the specific optimization formula is: F opt =F init κ; where F opt represents the optimized diagnosis result, F init This is the preliminary diagnosis result.

[0049] In a preferred embodiment, the operation process of the model optimization module includes the following contents:

[0050] According to the confirmed adjustment coefficient, the diagnostic model parameters of the local node are optimized; the optimized local node diagnostic model is deployed to the key electrical quantity and waveform data extraction task in the subsequent collected switch quantity information and fault recording information.

[0051] The technical effects and advantages of the power fault diagnosis and prediction system based on big data of the present invention are as follows:

[0052] 1. Through multi-level data preprocessing, feature extraction and dynamic pattern matching, localized intelligent diagnosis of power faults is realized. Combined with the adaptive update mechanism, the fault identification model is continuously optimized to improve the accuracy and reliability of diagnosis. Under limited transmission conditions, key electrical quantities and waveform data are uploaded first to ensure the stable operation and rapid recovery of the power system. Through the priority classification and sorting upload mechanism, combined with multi-dimensional comprehensive analysis and dynamic adjustment, the efficiency and accuracy of data transmission are improved, ensuring that the most relevant data is transmitted first in the event of sudden complex faults, reducing diagnostic errors caused by delays or losses, and improving the analysis efficiency of the central dispatching center and the stability and security of the system.

[0053] 2. The node stability fusion index calculated by combining the global deviation aggregation index and the trust stability harmonic measurement index at the central dispatching center realizes the accurate classification of the controllability of local nodes in the power system. This comprehensive judgment index can fully reflect the overall drift and stability of the node during the diagnosis process, so that the system can more efficiently identify the three types of controllable nodes: high, medium and low. In the case of medium controllable nodes, the diagnosis of local nodes is further optimized to ensure more accurate acquisition of key electrical quantities and waveform data, and reduce diagnostic errors by prioritizing the transmission of the most relevant data; the dynamic adjustment mechanism responds to fault evolution in real time, continuously optimizes the diagnostic results, and comprehensively improves the analysis efficiency of the central dispatching center and the stability and safety of the system, ensuring the efficient and stable operation of the power system under complex conditions, and providing strong technical support for intelligent diagnosis and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a structural schematic diagram of the power fault diagnosis and prediction system based on big data of the present invention. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] Example 1

[0057] Figure 1 The present invention provides a power fault diagnosis and prediction system based on big data, including:

[0058] Data acquisition module: obtains the switch quantity information and fault recording information corresponding to each local node, applies the local node diagnosis model to perform preliminary diagnosis, marks key electrical quantities and waveform data, and transmits the preliminary diagnosis results and the marked electrical quantities and waveform data to the data upload module;

[0059] Data upload module: Based on the preliminary diagnosis results, upload the marked key electrical quantities and waveform data to the central dispatch center and pass them to the data analysis module;

[0060] Data analysis module: receives key electrical quantities and waveform data, uses pre-built models for diagnosis, generates final diagnosis results, and transmits the results to the comparison analysis module and the fault prediction module;

[0061] Comparative analysis module: compares the preliminary diagnosis results with the final diagnosis results, extracts drift information and reliable information, records deviation data, builds a trust stability harmonic measurement index, classifies the controllability of local nodes, and divides them into three controllable categories: high, medium, and low, and passes the classification results to the adjustment optimization module;

[0062] Adjustment and optimization module: calculates the adjustment coefficient according to the trust stability harmonic measurement index, optimizes the preliminary diagnosis results of the medium controllable local nodes, and transmits the optimized diagnosis results to the model optimization module;

[0063] Model optimization module: Apply adjustment coefficients to optimize the local node diagnosis model, deploy the optimized local node diagnosis model to the key electrical quantity and waveform data extraction tasks in the subsequent collected switch quantity information and fault recording information, and send the optimized local node diagnosis model to the data acquisition module;

[0064] Fault prediction module: compare the final diagnosis results with historical data to extract features, conduct real-time monitoring and identification, calculate the fault probability, and generate early warning signals.

[0065] In modern power systems, the monitoring and fault diagnosis of power equipment are crucial, and local nodes and central dispatching centers play a key role in this process. Local nodes refer to independent monitoring and control devices distributed in the power network, which are responsible for real-time collection and processing of data related to the corresponding power equipment. These data include switch quantity information and fault recording information. Switch quantity information mainly records the state changes of power equipment, such as the opening and closing of circuit breakers, the operation of relays, etc., usually expressed in binary form, reflecting the working state of the equipment at a certain moment. Fault recording information refers to the detailed electrical quantity change data recorded by the fault recorder when an abnormality or failure occurs in the power system, such as the waveform of voltage and current. This information can provide the instantaneous state of the power system when the fault occurs, and help technicians deeply analyze the nature and cause of the fault.

[0066] The local node collects these switch quantity information and fault recording information to perform preliminary fault diagnosis and mark the key electrical quantities and important waveform data. Subsequently, these marked data are uploaded to the central dispatching center first. As the core analysis platform of the entire power system, the central dispatching center uses its full network data and pre-built complex models to conduct in-depth diagnosis of the key data uploaded by the local node, thereby generating more detailed and accurate final diagnosis results.

[0067] However, in the prior art, transmission delays are often encountered in the process of uploading fault recording files from local nodes to the central dispatching center. Since the recording files contain a large amount of waveform data, the transmission delay may not only cause the central dispatching center to be unable to obtain complete fault information in a timely manner, but also reduce the accuracy of fault diagnosis. The existence of transmission delays means that in an emergency situation where a fault occurs, the central dispatching center may not be able to quickly and accurately analyze it, affecting the rapid recovery of the power system and the efficiency of fault handling. Therefore, in practical applications, there is an urgent need for a fault diagnosis solution that can effectively deal with the problem of transmission delays, ensure accurate acquisition of key fault information in the shortest time, and improve the diagnostic efficiency and reliability of the overall system.

[0068] The deficiency of existing technologies is that they fail to fully deal with the negative impact of transmission delay on fault diagnosis, especially the lack of optimization means in the priority transmission and comprehensive analysis of key data. Therefore, it is particularly important to propose a processing solution that can not only perform efficient preliminary diagnosis at the local node level, but also quickly analyze the key data transmitted by priority at the central dispatching center to improve the accuracy and timeliness of fault diagnosis. Through this solution, the diagnostic deviation caused by transmission delay can be effectively reduced, and the power system's ability to respond to sudden faults can be enhanced.

[0069] In the power system, rapid identification and response of faults are crucial, especially in large-scale power networks, where any delay may lead to serious system interruption and economic losses. Traditional power fault diagnosis relies on the unified analysis of network-wide data by the central dispatching center, but this method has problems of data transmission delay and slow response speed, especially in large-scale networks. With the development of smart grid technology, each local node has gradually acquired the ability to collect and conduct preliminary data analysis autonomously. The data acquisition module realizes the preliminary identification of fault modes by collecting switch quantity information and fault recording information in real time, combined with multi-level data preprocessing and feature extraction technology. This localized fault identification method can not only obtain key electrical quantity information and waveform data at the first time of the fault, but also perform real-time matching and adaptive update through intelligent algorithms, which significantly improves the accuracy and response speed of fault identification, and provides high-quality data support for subsequent in-depth analysis and central dispatching.

[0070] The operation process of the data acquisition module includes the following:

[0071] A1, each local node continuously collects the switch quantity information and fault recording information related to the corresponding power equipment in real time. The switch quantity information includes the state change data of the circuit breaker, relay and other equipment, and records the switch state of the equipment. The fault recording information includes the instantaneous waveform data of electrical parameters such as voltage and current, which is automatically triggered and recorded by the fault recorder when the equipment is abnormal or fails.

[0072] A2, after collecting the switch quantity information and fault recording information, the local node immediately performs multi-level data preprocessing. First, the switch quantity information is time synchronized to ensure that the timestamp is consistent with the fault recording information. Then, the fault recording information is denoised and feature extracted. The noise in the signal is removed by the filter, and the key waveform data related to the fault characteristics is extracted, such as voltage mutation, current overload and other feature points. This step ensures that the processed data is highly consistent in the time domain, which is convenient for subsequent analysis.

[0073] A3, based on the preprocessed data, firstly, perform detailed feature extraction on the preprocessed switch quantity information and fault recording information. For the switch quantity information, extract the key state change points, such as the opening and closing time of the circuit breaker, the action time of the relay, etc. For the fault recording information, perform multi-scale analysis on the waveform through wavelet transform to extract the mutation points, oscillation frequency and harmonic components of electrical quantities such as current and voltage in different time periods. These features will serve as the basic data input for identifying fault modes.

[0074] For the switch quantity information and fault recording information collected from different sources, multi-channel data fusion technology is used to fuse the feature data of the two. Specifically, the switch quantity information is aligned with the fault recording data of the corresponding time period, and the features from these different sources are integrated into a unified feature vector through the weighted fusion method. The weight allocation is based on the importance analysis of the historical fault data. For example, in some cases, the switch quantity information may be more critical, while in other cases, the fault recording data may dominate. Through fusion, these feature vectors can more comprehensively reflect the overall characteristics of the fault.

[0075] After the fusion feature vector is generated, the dynamic pattern matching algorithm is used to match these features with the pre-stored fault pattern library in real time. The pattern library stores standard feature templates of various known fault types, which are established through historical data and expert experience. Dynamic pattern matching not only considers the absolute value of the feature, but also captures the dynamic trend of feature changes through time series analysis. The use of the Dynamic Time Warping (DTW) algorithm allows flexible matching on the time axis to adapt to the time differences and nonlinear changes that may exist in actual faults. This process can improve the accuracy of fault identification, especially when facing complex fault modes.

[0076] The process of fault pattern recognition is not static, but is continuously optimized through an adaptive update mechanism. After each fault pattern match is completed, the matching result is compared with the subsequent situation of the actual fault. If the matching effect is found to be unsatisfactory or there is a significant error, the system will automatically adjust the corresponding template parameters in the pattern library to better adapt it to the actual situation. The specific update process includes adjusting feature weights, correcting time series matching strategies, and adding new fault templates when necessary. This adaptive update mechanism enables the fault pattern library to gradually improve its matching accuracy and coverage as time goes by and new data accumulates.

[0077] Based on the real-time matching results, the local node can preliminarily identify the possible fault type (such as short circuit, ground fault, overload, etc.) and the time period when the fault occurs, thereby obtaining a preliminary diagnosis result.

[0078] A4. After obtaining the preliminary diagnosis results, the local node marks the key electrical quantity information and waveform data related to the fault. The marked electrical quantity information usually includes parameters with large changes before and after the fault, such as significant fluctuations in current and voltage. The marked waveform data includes the time period where the characteristic points are located, such as the key time window waveform before and after the fault. This step ensures that the most important data can be processed and transmitted first during the data transmission process.

[0079] The data acquisition module realizes the localization and intelligent preliminary diagnosis of power faults by introducing multi-level data preprocessing, feature extraction and dynamic pattern matching technology. Compared with the traditional fault diagnosis method that relies on central dispatch, the data acquisition module can collect and process key data in real time, greatly shortening the fault response time and reducing the diagnostic errors caused by data transmission delays. The addition of the adaptive update mechanism enables the fault identification model to be continuously optimized with the accumulation of time and data, improving the accuracy and reliability of diagnosis. In addition, by marking and prioritizing key electrical quantities and waveform data, the data acquisition module ensures that the most important information can be uploaded first under limited transmission conditions, ensuring the stable operation and rapid recovery of the power system. This module has significant practical application value in improving fault diagnosis efficiency and reducing system failure risks.

[0080] In the fault diagnosis process of the power system, the timeliness and accuracy of the data are crucial, especially when sudden faults occur. Rapid acquisition and transmission of key data can significantly improve the efficiency and effectiveness of fault handling. The data upload module implements data priority classification and adopts an orderly upload strategy by performing importance analysis, correlation evaluation and fault type matching on the marked electrical quantity information and waveform data. The core of this module is to ensure that the most valuable and representative data can be uploaded to the central dispatching center first when the network transmission conditions are unstable or the bandwidth is limited, thereby supporting more accurate fault diagnosis and analysis. With the continuous development of smart grids, the amount and complexity of data in power systems are increasing. The data upload module ensures the efficient transmission of key data by optimizing the data transmission order and priority classification, avoiding diagnostic lags caused by transmission delays, and providing reliable guarantees for the stable operation of the power grid.

[0081] The operation process of the data upload module includes the following:

[0082] B1, after the initial diagnosis is completed, the importance of the marked electrical quantity information and waveform data is first analyzed. By comprehensively evaluating the timestamp, amplitude change, frequency characteristics, etc. of the data, the most representative and most significantly changed parameters before and after the fault occurs are identified. For example, the sudden change point, duration, and degree of deviation from historical data of current and voltage are analyzed first. These analysis results will serve as the basis for data priority classification.

[0083] B2, Next, the correlation between the marked electrical quantity information and waveform data is evaluated. Through correlation analysis and feature clustering, electrical quantities and waveform features that are highly correlated with each other are identified. For example, if the current and voltage waveforms at a certain moment change dramatically and synchronously, the two parameters are considered to be highly correlated and belong to the same priority. This correlation evaluation ensures that closely related data can be transmitted together when data is uploaded, facilitating a more complete fault analysis by the central dispatch center.

[0084] B3, according to the fault type identified in the preliminary diagnosis, the marked electrical quantity information and waveform data are matched. The marked electrical quantity information and waveform data are compared with the preset fault type database to determine which data is most critical in the analysis of this type of fault. For example, in a short circuit fault, the waveform feature of the current mutation may have the highest priority, while in a ground fault, the voltage drop feature may be given priority. This matching process ensures that the most critical data of each fault type can be uploaded first.

[0085] B4, based on the analysis results of the above steps, a priority score is assigned to each set of marked electrical quantity information and waveform data. The score is calculated based on the three dimensions of data importance, relevance, and fault type matching. The calculation formula can be designed as a weighted sum model, in which the weight of each dimension is dynamically adjusted according to historical fault data. For example, a dimension with a higher fault type matching degree can be given a higher weight to ensure that the key data of the fault type is uploaded first.

[0086] B5, sort the data from the highest priority to the lowest priority. The electrical quantity information and waveform data ranked first are marked as key electrical quantity information and waveform data, and are immediately prepared for upload. They are given priority when network bandwidth is limited, and other data are uploaded gradually. The data ranked later are uploaded last or delayed when necessary. This sorted upload storage ensures that the most important data is processed first under any transmission conditions.

[0087] The data upload module significantly improves the efficiency and accuracy of data transmission through a sophisticated priority classification and sorting upload mechanism. The multi-dimensional comprehensive analysis and dynamic adjustment strategy adopted by the module ensures that the most critical electrical quantity information and waveform data can be transmitted first during each fault handling process, reducing fault diagnosis errors caused by data delays or losses. By combining the relevance of the marked data with the fault type matching, the system can more specifically transmit the data most relevant to the current fault, further improving the analysis efficiency of the central dispatching center. Especially in the face of sudden and complex power failures, the module's innovative processing method can effectively deal with the uncertainty of transmission conditions and ensure the stability and safety of the power system.

[0088] In modern power systems, accurate fault diagnosis is crucial to the safety and reliability of the system. With the development of smart grids, the amount and complexity of data generated by power systems are increasing. How to quickly and accurately identify faults in massive data has become a major challenge. As an important part of the central dispatching center, the data analysis module provides a comprehensive fault diagnosis platform by efficiently and accurately analyzing key electrical quantity information and waveform data uploaded by local nodes. This module can not only process diverse data from different nodes, but also ensure the integrity and accuracy of data during transmission through multiple processing steps such as standardization, feature reconstruction, model optimization and dynamic adjustment. Through parallel computing and model integration, the data analysis module can quickly process complex power fault scenarios and generate high-precision diagnostic results, providing strong support for the stable operation of the power system.

[0089] The operation process of the data analysis module includes the following:

[0090] C1, after receiving the key electrical quantity information and waveform data uploaded by the local node, the central dispatch center first performs data integrity verification. Use the pre-generated verification code to verify each segment of data to ensure that the data is not damaged or lost during the transmission process. If the verification fails, it will request retransmission or use the data copy of the backup path for replacement.

[0091] C2, after verification, standardizes the received key data. Since data from different local nodes may have differences in format, unit, and accuracy, the central dispatch center uses a unified standardization algorithm to process all data. This process includes unit conversion, data format conversion, and synchronous processing of data at different time scales to ensure that the data can fully match the input requirements of the pre-built model.

[0092] C3, to improve diagnostic accuracy, reconstruct and enhance the features of the standardized data. Through advanced feature extraction algorithms, such as principal component analysis (PCA) or nonlinear feature mapping, key electrical quantity information and waveform data are converted into a higher-dimensional feature space. This process can extract potential patterns and relationships in the data, providing richer input features for subsequent model diagnosis. Feature enhancement also includes spectral analysis of waveform data to capture potential harmonics and frequency changes, which are often important indicators in fault analysis.

[0093] C4, the central dispatch center selects the most suitable pre-built model for diagnosis based on the preliminary diagnosis results of the fault type. Multiple pre-built models are stored, and each model is optimized for different types of faults. During the model selection process, the parameters of the model are dynamically tuned based on the characteristics of the received key data and historical fault data to better adapt to the current data characteristics. For example, in the short-circuit fault analysis, the model for current mutation will be given priority, and the response threshold of the model will be adjusted to improve the recognition accuracy.

[0094] C5, to improve diagnostic efficiency and accuracy, the central dispatch center runs the selected pre-built models in parallel on multiple computing nodes. Each node processes data independently and generates corresponding diagnostic results. After that, the results of multiple nodes are integrated into the model, and the outputs of each model are integrated using weighted voting, ensemble learning and other methods to generate a more accurate final diagnostic result. This process takes advantage of the advantages of multiple models and reduces the deviations and errors that may exist in a single model.

[0095] C6, after generating the preliminary final diagnosis results, dynamically adjusts the diagnosis results based on real-time monitoring data and feedback from other nodes in the central dispatch center. If new information is found during the subsequent data reception process, the input data will be automatically updated and the final diagnosis results will be recalculated. This dynamic adjustment mechanism ensures the real-time and accuracy of the diagnosis results, especially when the fault type is complex or the fault is evolving, it can quickly adjust and give the optimal solution.

[0096] C7, when the final diagnosis result is determined, a complete diagnosis report is generated. The report content includes the specific type of fault, the time period of occurrence, the equipment or area affected, and the recommended treatment measures. These contents correspond to the preliminary diagnosis report one by one to ensure that the difference between the preliminary diagnosis result and the final diagnosis result can be compared later. In addition, the report also contains the detailed process of data analysis, the log of model operation, and the detailed description of feature reconstruction and model selection for subsequent review and optimization. Through this correspondence, the accuracy of the diagnosis process and the effect of model optimization can be effectively verified, providing a reliable basis for the continuous improvement of the system.

[0097] C8, after the final diagnosis report is generated, the results are compared and verified with historical data to ensure the accuracy of the diagnosis. If the result deviation is large, the relevant information of this diagnosis will be recorded and archived for use in subsequent model training and optimization processes. At the same time, the system will store the diagnosis results in the central database as reference data for future fault analysis and trend prediction.

[0098] The data analysis module achieves high efficiency and accuracy in power fault diagnosis through a series of innovative data processing and diagnosis technologies. First, the data integrity check and standardization processing steps ensure that data from different local nodes can be analyzed on a unified platform, effectively eliminating potential errors caused by data format and unit differences. The feature reconstruction and enhancement steps extract deep patterns in the data through advanced algorithms, significantly improving the input quality of the diagnosis model. The model selection and dynamic tuning mechanism enable the system to flexibly adapt to different types of fault characteristics and ensure the accuracy of the diagnosis results. Parallel computing and model integration take advantage of the advantages of multiple models, reduce the possible deviations of a single model, and further improve the reliability of diagnosis. Through the dynamic adjustment mechanism, the system can respond to the evolution of faults in real time and provide continuously optimized diagnostic results. Finally, the one-to-one correspondence between the detailed diagnostic report and the preliminary diagnostic results not only improves the transparency of the diagnosis, but also provides a solid data foundation for subsequent system optimization and fault trend prediction. The overall design of this module provides a strong technical guarantee for the intelligent diagnosis and maintenance of the power system.

[0099] In complex power systems, accurate evaluation of the controllability of local nodes is crucial to ensuring the stability of the power system. Each node in the power system may exhibit different characteristic drifts and stability differences when a fault occurs. Traditional single indicator evaluation methods are often difficult to fully reflect the overall status of the node. In order to overcome this challenge, the comprehensive use of the "global deviation aggregation index" and the "trust stability harmonic measurement index" provides a more refined and comprehensive evaluation method by constructing a comprehensive judgment index.

[0100] The "Global Deviation Aggregation Index" is used to quantify the overall deviation of the diagnostic results in the entire domain, reflecting the degree of change of the system at each feature level. Through this index, the difference between the diagnostic results and the actual fault can be accurately captured, thereby evaluating the drift of the system. The "Trust Stability Harmonic Measurement Index" focuses on the drift direction and stability of the system in the time series, and provides a comprehensive measure of the stability and reliability of the system by synthesizing the drift direction vectors at different time points.

[0101] By combining these two indexes, the calculated "node stability fusion index" can more accurately reflect the overall controllability of the node. This comprehensive judgment index not only takes into account the drift of the system, but also associates it with reliability, thus providing a scientific basis for the classification and priority setting of nodes. Through this comprehensive evaluation, the three types of controllable nodes, high, medium and low, can be effectively distinguished, ensuring that more accurate and efficient measures are taken in the process of power system fault management and resource allocation.

[0102] The operation process of the comparative analysis module includes the following:

[0103] Drift information and reliable information between the preliminary diagnosis result and the final diagnosis result are extracted and compared, wherein the drift information includes the global deviation aggregation index, and the reliable information includes the trust stability harmonic measurement index; the global deviation aggregation index and the trust stability harmonic measurement index are comprehensively processed and calculated to obtain the trust stability harmonic measurement index, for example, it can be calculated using the following formula: Among them, IC represents the trust stability harmonic measurement index, η1 and η2 are preset proportional coefficients, which control the weights of the two indexes in the comprehensive judgment. The purpose is to combine high reliability (i.e., large trust stability harmonic measurement index value) with low drift (i.e., small global deviation aggregation index value) to accurately reflect the controllability of the node.

[0104] The trust stability harmonic index is used to represent and reflect the overall stability and reliability of the system during the fault diagnosis process. The index combines the drift direction and cumulative drift of the system at different times, and measures the deviation and consistency between the diagnosis results and the actual fault performance through the analysis of time series data. The larger the value of the trust stability harmonic index, the higher the stability and reliability of the system diagnosis results, and the more consistent and reliable the system's performance is when facing complex faults; the smaller the value, the lower the stability of the system diagnosis results, the greater the drift or uncertainty, and the inconsistent performance of the system at different time points, which may require further optimization and adjustment.

[0105] According to the value of the trust stability harmonic measurement index, the controllability of the local node is classified. The classification criteria are as follows:

[0106] High controllability: When the value of the trust stability harmonic index is less than the classification threshold of one, the node is considered to have high controllability.

[0107] Moderate controllability: When the value of the trust stability harmonic index is greater than or equal to the classification threshold one and less than the classification threshold two, the node is considered to have moderate controllability.

[0108] Low controllability: When the value of the trust stability harmonic index is greater than or equal to the classification threshold 2, the node is considered to have low controllability.

[0109] The node stability fusion index calculated by combining the global deviation aggregation index and the trust stability harmonic measurement index realizes the accurate classification of the controllability of local nodes in the power system. This comprehensive judgment index can fully reflect the overall drift and stability of the node during the diagnosis process, allowing the system to more efficiently identify high, medium and low controllable nodes. This evaluation method not only improves the system's ability to respond to complex faults, but also optimizes resource allocation, ensuring that the power system maintains efficient and stable operation under various complex situations, and has significant application value.

[0110] The generation logic and detailed steps of the global deviation aggregation index are as follows:

[0111] D1, first, construct a drift feature vector to quantify the difference between the preliminary diagnosis result and the final diagnosis result. Set the drift feature vector D, which contains drift information in multiple dimensions, each dimension corresponding to a feature in the diagnosis result. For the i-th feature, calculate its drift value d i : in, represents the i-th eigenvalue of the final diagnosis result, The i-th feature value represents the preliminary diagnosis result. The drift value represents the degree of change of the feature between the preliminary diagnosis and the final diagnosis.

[0112] D2, in order to evaluate the reliability of each feature drift value, the fuzzy correlation function μ(d i ), which is used to measure the rationality of the drift value relative to the historical data. The fuzzy correlation function is defined as: Among them, λ i is the threshold parameter set for the i-th feature based on historical data, indicating the reasonable range of variation of the feature; α is the first adjustment parameter, which controls the steepness of the function and determines the sensitivity of the fuzzy association to the degree of deviation. The value range of the fuzzy membership is [0,1]. The closer the value is to 1, the more reasonable the drift value is.

[0113] D3, according to the fuzzy membership, assign weight w to each feature i The weight calculation formula is: Among them, γ is the second adjustment parameter, which controls the influence of drift value on weight; S i Score the importance of the i-th feature; S j Score the importance of the jth feature. Through this formula, the rationality of the drift value and the importance of the feature are combined to ensure that the key features have a greater weight in the reliability assessment.

[0114] The importance score is obtained by combining statistical analysis of historical fault data with expert experience. First, for a large amount of historical data, the performance of each feature in different fault types is quantitatively analyzed, including its change range, stability, and degree of correlation with other features before and after the fault occurs. Through these data analysis, the sensitivity and contribution of each feature in different fault scenarios can be calculated, so as to preliminarily determine its importance score. At the same time, in order to ensure the accuracy and practicality of the score, expert experience is also combined. Especially in specific scenarios, experts can adjust or fine-tune the importance of certain features based on their deep understanding of the system. This scoring method not only relies on objective data analysis, but also fully considers the empirical judgment in practical applications to ensure that the scoring results can accurately reflect the role of features in fault diagnosis. Therefore, the process of obtaining the importance score is based on the organic combination of data-driven and empirical knowledge. Such a setting not only ensures the scientific nature of the score, but also enhances its applicability in actual fault diagnosis and avoids the risk of insufficient disclosure.

[0115] D4, combine the drift value of each feature with its corresponding weight to calculate the global deviation aggregation index R of the overall diagnosis result. The calculation formula is: Among them, ∈ is a small constant used to avoid the denominator being zero; n represents the number of features. The value range of the global deviation aggregation index is [0,1]. The higher the value, the stronger the reliability of the diagnosis result.

[0116] The generation logic and detailed steps of the trust stability harmonic index:

[0117] E1, first, the performance of each key feature of the preliminary diagnosis results and the final diagnosis results at different times is calculated to construct the drift matrix Q. Each element q in the matrix x,y Indicates the relative drift of feature x at time point y. The drift is calculated by dividing the difference between the final diagnosis result and the preliminary diagnosis result by the reference value. In this way, the relative drift of each feature at different time points is obtained.

[0118] Construct the drift matrix Q, each element q in the matrix x,y It represents the relative drift of feature x at time point y. The drift value calculation formula is: in, Represents the value of feature x at time point y in the final diagnosis result, Indicates the corresponding value in the preliminary diagnosis result, is the reference value of the corresponding feature.

[0119] E2, after obtaining the drift matrix, calculate the overall drift accumulation function based on the drift matrix. The overall drift accumulation function is obtained by performing nonlinear transformation on the drift values ​​of all features and summing them at the time point y. The role of the nonlinear transformation function is to emphasize or weaken the influence of the drift amount so that the overall drift accumulation function reflects the overall drift degree of the system at a specific time point.

[0120] Based on the drift matrix, the overall drift accumulation function R(y) is calculated to calculate the cumulative drift value of all features at time y. The formula is as follows: Among them, φ(q x,y ) is the nonlinear conversion function of the drift value, which is in the form of Where ξ is the third adjustment parameter; p is the total number of features.

[0121] E3, by judging the sign of the overall drift accumulation function and combining its absolute value, calculates the drift direction synthetic vector. The drift direction synthetic vector represents the drift direction and the accumulated drift amount over the entire time range. The sign function is used to identify the positive and negative directions of the drift, while the absolute value operation ensures the cumulative characteristics of the drift amount.

[0122] Calculate the directionality of the overall drift by constructing a synthetic drift direction vector Z to represent the drift direction over the entire time range. The formula is as follows: Where sgn(R(y)) is the sign function that determines the drift direction (positive or negative), |R(y)| is the absolute value of the accumulated drift, and T is the total time of fault evolution.

[0123] E4, finally, the trust stability harmonic index is calculated by combining the integral result of the drift direction synthetic vector and the overall drift accumulation function. By introducing the fourth adjustment parameter to control the influence of the cumulative drift on the trust stability harmonic index, the overall drift performance in the entire diagnostic process is obtained. The trust stability harmonic index reflects the overall deviation between the system diagnosis result and the actual fault condition, and is ultimately used to measure the stability and accuracy of the system.

[0124] Combined with the value of the overall drift accumulation function and the drift direction synthetic vector Z, the final trust stability harmonic measurement index V is calculated. The formula is as follows: in, is the fourth adjustment parameter, controlling the impact of cumulative drift on the trust stability harmonic measurement index; The integral of the accumulated drift is used to evaluate the amount of drift over the entire time period.

[0125] In the process of fault diagnosis and processing of power systems, the accuracy and reliability of node diagnostic models directly affect the overall stability of the system and the efficiency of fault response. However, traditional diagnostic models often have solidification problems and are difficult to flexibly respond to complex and changeable actual fault conditions. The adjustment optimization module and the model optimization module optimize the preliminary diagnostic results by introducing dynamic adjustment coefficients, and then dynamically adjust the diagnostic model according to the optimized results, thereby realizing adaptive optimization and real-time correction of the model. This method can continuously adjust the diagnostic model to the optimal state, so that it has higher accuracy and stability when processing real-time data. By deploying the optimized model to subsequent data processing tasks, it ensures that the system can continuously and reliably extract and analyze key electrical quantities and waveform data, providing solid technical support for the intelligent operation of the power system.

[0126] The operation process of the adjustment and optimization module includes the following:

[0127] F1, first, the trust stability harmonic index is calculated for the preliminary diagnosis results of the local node. Through the above method, the drift direction synthetic vector and the overall drift accumulation function are combined to obtain the trust stability harmonic index. This index reflects the stability and consistency of the node diagnosis results, and the specific value is used to judge whether the performance of the node at different times is sufficiently reliable.

[0128] F2, based on the value of the trust stability harmonic measurement index, preliminarily calculate the adjustment coefficient κ. The calculation formula of the adjustment coefficient is designed as: in, is the fifth adjustment parameter, which is used to control the sensitivity of the adjustment coefficient to the trust stability harmonic measurement index. ω is a nonlinear adjustment parameter, which ensures that when the trust stability harmonic measurement index is small, the adjustment coefficient can correct the preliminary diagnosis result to a large extent.

[0129] F3, for the preliminary diagnosis results of the medium controllable nodes, use the adjustment coefficient to optimize them. The optimization process adjusts the key parameters in the diagnosis model to make the final diagnosis results closer to the actual fault situation. The specific optimization formula is: F opt =F init κ; where F opt represents the optimized diagnosis result, F init This is the preliminary diagnosis result. The diagnosis result is corrected by adjusting the coefficient, so that the diagnosis model can show higher accuracy and stability on medium controllable nodes.

[0130] The operation process of the model optimization module includes the following:

[0131] According to the confirmed adjustment coefficient, the diagnostic model parameters of the local node are optimized. The diagnostic model usually includes multiple key parameters, such as fault threshold, feature extraction coefficient, filter setting, etc. For example, the application of the adjustment coefficient will adjust these parameters in the following way: P′=P·κ; where P represents the current model parameter and P′ is the adjusted model parameter. Through this adjustment, it is ensured that the model can better adapt to the actual fault characteristics and reduce errors and deviations when processing the current data. The optimized local node diagnostic model is deployed to the key electrical quantity and waveform data extraction tasks in the subsequent collected switch quantity information and fault recording information.

[0132] The adjustment coefficient is calculated based on the trust stability harmonic measurement index, and the diagnostic results of the local node are dynamically corrected to achieve more accurate fault identification on the medium controllable node. Further, by dynamically adjusting the parameters of the local node diagnostic model, it is ensured that the model classifies key electrical quantities and waveform data more accurately. Finally, the optimized model is deployed to subsequent data processing tasks to ensure the accuracy and data processing capabilities of the diagnosis during long-term operation. By improving the disconnection capacity of key electrical quantities and waveform data, more important information is transmitted first during data transmission, effectively reducing the diagnostic deviation caused by transmission delay, thereby greatly improving the accuracy of local fault diagnosis, enhancing the ability of the power system to respond to sudden failures, and providing a strong guarantee for the stability and security of the smart grid.

[0133] The operation process of the fault prediction module includes the following:

[0134] After obtaining the final diagnosis result generated by the central dispatch center, it is compared with the historical fault data, relevant features are extracted, and input into the fault prediction model. By continuously monitoring the real-time data of each local node, features similar to the final diagnosis result are identified, and the prediction model is applied to calculate the probability of potential faults. When the prediction model identifies a potential fault signal, an early warning notification is generated and corresponding preventive measures are recommended.

[0135] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0136] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0137] It should be noted that, in this article, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including a..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0138] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. The power fault diagnosis and prediction system based on big data is characterized by: Includes steps: Data acquisition module: obtains the switch quantity information and fault recording information corresponding to each local node, applies the local node diagnosis model to perform preliminary diagnosis, marks key electrical quantities and waveform data, and transmits the preliminary diagnosis results and the marked electrical quantities and waveform data to the data upload module; Data upload module: Based on the preliminary diagnosis results, upload the marked key electrical quantities and waveform data to the central dispatch center and pass them to the data analysis module; Data analysis module: receives key electrical quantities and waveform data, uses pre-built models for diagnosis, generates final diagnosis results, and transmits the results to the comparison analysis module and the fault prediction module; Comparative analysis module: compares the preliminary diagnosis results with the final diagnosis results, extracts drift information and reliable information, records deviation data, builds a trust stability harmonic measurement index, classifies the controllability of local nodes, and divides them into three controllable categories: high, medium, and low, and passes the classification results to the adjustment optimization module; Adjustment and optimization module: calculates the adjustment coefficient according to the trust stability harmonic measurement index, optimizes the preliminary diagnosis results of the medium controllable local nodes, and transmits the optimized diagnosis results to the model optimization module; Model optimization module: Apply adjustment coefficients to optimize the local node diagnosis model, deploy the optimized local node diagnosis model to the key electrical quantity and waveform data extraction tasks in the subsequent collected switch quantity information and fault recording information, and send the optimized local node diagnosis model to the data acquisition module; Fault prediction module: compare the final diagnosis results with historical data to extract features, conduct real-time monitoring and identification, calculate the fault probability, and generate early warning signals.

2. The power fault diagnosis and prediction system based on big data according to claim 1 is characterized by: The operation process of the data acquisition module includes the following: A1, each local node continuously collects the switch quantity information and fault recording information related to the corresponding power equipment in real time; A2, after collecting the switch quantity information and fault recording information, the local node immediately performs multi-level data preprocessing; A3, based on the preprocessed data, firstly extract the features of the preprocessed switch quantity information and fault recording information; adopt multi-channel data fusion technology to fuse the feature data of the switch quantity information and fault recording information; after the fusion feature vector is generated, use the dynamic pattern matching algorithm to match the features of the switch quantity information and fault recording information with the pre-stored fault pattern library in real time; based on the real-time matching results, the local node preliminarily identifies the fault type and the time period when the fault occurs, thereby obtaining a preliminary diagnosis result; A4, after obtaining the preliminary diagnosis results, the local node marks the key electrical quantity information and waveform data related to the fault.

3. The power fault diagnosis and prediction system based on big data according to claim 2 is characterized in that: The operation process of the data upload module includes the following: B1, after the preliminary diagnosis is completed, firstly, the importance analysis of the marked electrical quantity information and waveform data is carried out; B2, Next, the correlation between the marked electrical quantity information and the waveform data is evaluated; through correlation analysis and feature clustering, the electrical quantities and waveform features that are highly correlated with each other are identified; B3, matching the marked electrical quantity information and waveform data according to the fault type identified in the preliminary diagnosis; Compare the marked electrical quantity information and waveform data with the preset fault type database to determine which data is most critical in the corresponding type of fault analysis; B4, based on the analysis results of the above steps, assign a priority score to each group of marked electrical quantity information and waveform data; B5, sort the results from the highest priority to the lowest priority; The electrical quantity information and waveform data ranked first are marked as key electrical quantity information and waveform data and are scheduled to be uploaded immediately.

4. The power fault diagnosis and prediction system based on big data according to claim 3 is characterized in that: The operation process of the data analysis module includes the following: C1, after the central dispatch center receives the key electrical quantity information and waveform data uploaded by the local node, it first performs data integrity verification; C2, after verification, standardize it; C3, feature reconstruction and enhancement after standardization; C4, the central dispatch center selects the corresponding pre-built model for diagnosis based on the preliminary diagnosis results of the fault type; C5, the central dispatch center runs the selected pre-built models in parallel on multiple computing nodes; C6, after generating the preliminary diagnosis results, dynamically adjusts the diagnosis results based on the real-time monitoring data and feedback from other nodes in the central dispatch center to determine the final diagnosis results.

5. The power fault diagnosis and prediction system based on big data according to claim 4 is characterized in that: The operation process of the comparative analysis module includes the following: Extract and compare the drift information and reliable information between the preliminary diagnosis results and the final diagnosis results, where the drift information includes the global deviation aggregation index and the reliable information includes the trust stability harmonic measurement index; the global deviation aggregation index and the trust stability harmonic measurement index are comprehensively processed and calculated to obtain the trust stability harmonic measurement index; according to the value of the trust stability harmonic measurement index, the controllability of the local node is classified; the classification criteria are as follows: When the value of the trust stability harmonic index is less than the classification threshold one, the node is considered to have high controllability; when the value of the trust stability harmonic index is greater than or equal to the classification threshold one and less than the classification threshold two, the node is considered to have medium controllability; when the value of the trust stability harmonic index is greater than or equal to the classification threshold two, the node is considered to have low controllability.

6. The power fault diagnosis and prediction system based on big data according to claim 5 is characterized in that: The generation logic and detailed steps of the global deviation aggregation index are as follows: D1, first, construct a drift feature vector to quantify the difference between the preliminary diagnosis result and the final diagnosis result; set the drift feature vector D, which contains drift information in multiple dimensions, each dimension corresponds to a feature in the diagnosis result; for the i-th feature, calculate its drift value d i : in, represents the i-th eigenvalue of the final diagnosis result, The i-th eigenvalue represents the preliminary diagnosis result; D2, to evaluate the reliability of each feature drift value, the fuzzy correlation function μ(d i ), which is used to measure the rationality of the drift value relative to the historical data; the fuzzy correlation function is defined as: Among them, λ i is the threshold parameter set for the i-th feature based on historical data; α is the first adjustment parameter, which controls the steepness of the function; D3, according to the fuzzy membership, assign weight w to each feature i ; The weight calculation formula is: Among them, γ is the second adjustment parameter, which controls the influence of drift value on weight; S i Score the importance of the i-th feature; S j Score the importance of the jth feature; D4, combine the drift value of each feature with its corresponding weight to calculate the global deviation aggregation index R of the overall diagnosis result; the calculation formula is: Among them, ∈ is a small constant used to avoid the denominator being zero; n represents the number of features.

7. The power fault diagnosis and prediction system based on big data according to claim 5 is characterized in that: The generation logic and detailed steps of the trust stability harmonic index: E1, construct the drift matrix Q, each element q in the matrix x,y Indicates the relative drift of feature x at time point y; the drift value calculation formula is: in, Represents the value of feature x at time point y in the final diagnosis result; Indicates the corresponding value in the preliminary diagnosis result; is the reference value of the corresponding feature; E2, based on the drift matrix, calculates the overall drift accumulation function R(y), which is used to calculate the cumulative drift value of all features at time y; the formula is as follows: Among them, φ(q x,y ) is the nonlinear conversion function of the drift value, which is in the form of Where ξ is the third adjustment parameter; p is the total number of features; E3, calculates the directionality of the overall drift by constructing a drift direction synthetic vector Z, which represents the drift direction within the entire time range; the formula is as follows: Where sgn(R(y)) is the sign function to determine the drift direction (positive or negative), |R(y)| is the absolute value of the accumulated drift; T is the total time of fault evolution; E4, combining the value of the overall drift accumulation function and the drift direction synthetic vector Z, calculates the final trust stability harmonic measurement index V; the formula is as follows: in, is the fourth adjustment parameter, controlling the impact of cumulative drift on the trust stability harmonic measurement index; The integral of the accumulated drift is used to evaluate the amount of drift over the entire time period.

8. The power fault diagnosis and prediction system based on big data according to claim 1 is characterized by: The operation process of the adjustment and optimization module includes the following: F1, first, calculate the trust stability harmonic index of the preliminary diagnosis results of the local node; F2, based on the value of the trust stability harmonic index, preliminarily calculate the adjustment coefficient k; The calculation formula of the adjustment coefficient is designed as: in, is the fifth adjustment parameter, used to control the sensitivity of the adjustment coefficient to the trust stability harmonic measurement index, and ω is the nonlinear adjustment parameter; F3, for the preliminary diagnosis results of the medium controllable nodes, the adjustment coefficient is used to optimize them; the optimization process adjusts the key parameters in the diagnosis model to make the final diagnosis results closer to the actual fault situation; the specific optimization formula is: F opt =F init k; where F opt represents the optimized diagnosis result, F init This is the preliminary diagnosis result.

9. The power fault diagnosis and prediction system based on big data according to claim 8 is characterized in that: The operation process of the model optimization module includes the following: According to the confirmed adjustment coefficient, the diagnostic model parameters of the local node are optimized; the optimized local node diagnostic model is deployed to the key electrical quantity and waveform data extraction task in the subsequent collected switch quantity information and fault recording information.

Images