A wind farm high-voltage circuit breaker fault detection method and system

Abstract

The application discloses a kind of wind farm high-voltage circuit breaker fault detection method, system, electronic equipment and storage medium, belong to electric power equipment fault detection technical field.By collecting circuit breaker multi-source sensing data and wind farm operating condition data, construct dynamic operating condition benchmark domain to filter operating condition disturbance interference, output effective abnormal signal.Through extracting multidimensional fault feature, complete abnormal preliminary judgment, and introduce multi-level confidence score to realize detection result quantization correction.Relying on fault correlation matrix and using progressive stripping strategy, decouple coupled features, separate independent fault features.Combined with historical time series data, trace back fault degradation transmission path, predict fault evolution trend.At the same time, through double feedback channel iteration updates fault feature library and correlation matrix.The application can effectively reduce fault false alarm rate, accurately locate fault source, realize fault prediction, adapt to complex dynamic operating condition of wind farm, and can provide reliable technical support for circuit breaker predictive maintenance and power grid safety and stability operation.

Description

Technical Field

[0001] This invention relates to the field of power system fault diagnosis technology, and in particular to a method, system, electronic equipment, and computer storage medium for detecting faults in high-voltage circuit breakers in wind farms based on multi-level confidence transfer and dynamic operating condition benchmarks. Background Technology

[0002] In wind farms, high-voltage circuit breakers are key equipment for ensuring the safe operation of the power grid. Because wind farms are usually located in harsh environments such as deserts, mountains, or offshore, and the output of wind turbines is significantly fluctuating and intermittent, high-voltage circuit breakers are subjected to frequent operation and complex stress conditions for a long time.

[0003] Existing circuit breaker fault detection technologies mainly suffer from the following problems: 1. Poor adaptability to operating conditions: Traditional methods usually use fixed thresholds for alarms, which makes it difficult to distinguish between normal signal disturbances caused by sudden changes in wind speed and power fluctuations and real fault characteristics, resulting in a high false alarm rate.

[0004] 2. Severe coupling of fault characteristics: When a circuit breaker experiences a complex fault, electrical signals, mechanical vibration signals, and insulation signals often overlap. Existing technologies lack effective decoupling mechanisms, making it difficult to accurately locate the source of the fault.

[0005] 3. Lack of evolution prediction: Existing detection methods are mostly post-event alarms or simple status judgments, lacking the ability to predict the trend of fault deterioration, and thus failing to provide effective support for predictive maintenance.

[0006] Therefore, there is an urgent need for a detection method that can adapt to the dynamic operating conditions of wind farms, accurately decouple complex fault characteristics, and predict evolution trends. Summary of the Invention

[0007] To address the shortcomings of the existing technologies, this invention proposes a method, system, electronic equipment, and computer storage medium for fault detection of high-voltage circuit breakers in wind farms based on multi-level confidence transfer and dynamic operating condition benchmarks.

[0008] The technical solution adopted in this invention is as follows: In one implementation of the first aspect, this application provides a method for fault detection of high-voltage circuit breakers in wind farms, comprising the following steps performed in sequence: Step S1: Acquire multi-source raw sensing data of the high-voltage circuit breaker and operating condition data reflecting the operating status of the wind farm; construct a dynamic operating condition reference domain based on the operating condition data, and compare the multi-source raw sensing data with the dynamic operating condition reference domain to filter out interference signals caused by operating condition disturbances, and output a valid abnormal signal when it is determined that there is a continuous characteristic offset. Step S2: In response to receiving the valid abnormal signal, extract its multi-dimensional fault features and make a preliminary judgment on the abnormality category based on the basic fault feature library; Step S3: In response to the preliminary judgment result that there is a stable anomaly, the preset fault correlation matrix is ​​invoked to perform coupling and decoupling analysis on the multi-dimensional fault features corresponding to the stable anomaly, so as to separate the mutually superimposed independent fault features. Step S4: Based on the independent fault characteristics and the historical state time series data of the high-voltage circuit breaker, deduce the fault deterioration propagation path and predict the fault evolution trend.

[0009] In one implementation of the second aspect, this application provides a wind farm high-voltage circuit breaker fault detection system for implementing the method provided in the first aspect or any possible implementation of the first aspect, comprising: The operating condition disturbance filtering module is used to perform step S1; The explicit anomaly initial judgment module is communicatively connected to the operating condition disturbance filtering module and is used to execute step S2; The coupling fault decoupling analysis module is communicatively connected to the explicit anomaly preliminary judgment module and is used to execute step S3; The fault tracing and evolution prediction module is communicatively connected to the coupled fault decoupling analysis module and is used to execute step S4.

[0010] In one implementation of the third aspect, this application provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the program to implement the method provided by the first aspect or any possible implementation of the first aspect.

[0011] In one implementation of the fourth aspect, this application provides a computer-readable storage medium having a computer program stored thereon, characterized in that the program, when executed by a processor, implements the method provided by the first aspect or any possible implementation of the first aspect.

[0012] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: This invention addresses the industry pain points of high-voltage circuit breakers in wind farms, including strong fluctuations in operating conditions, coupled fault features, and lack of trend prediction, possessing multiple core technological advantages. It abandons the traditional fixed-threshold detection mode, constructing a dynamic operating condition reference domain adapted to wind speed and turbine output, effectively filtering out operating condition disturbances and significantly reducing false alarm rates. Simultaneously, it establishes a progressive decoupling mechanism for electrical, mechanical, and insulation features, combined with a fault correlation matrix to achieve precise separation of coupled fault features, solving the problem of ambiguous fault location in traditional technologies. Furthermore, this invention introduces a multi-level confidence quantification correction mechanism to improve the accuracy of the entire detection process, and uses time-series data to trace and deduce fault degradation paths, achieving fault trend prediction. Relying on a dual-feedback closed-loop iterative optimization mechanism, it continuously updates the feature library and correlation matrix, constantly improving detection adaptability, providing precise support for predictive maintenance of wind farm circuit breakers, and ensuring the stable operation of wind power equipment and the power grid. Attached Figure Description

[0013] The present invention will be described by way of example and with reference to the accompanying drawings, wherein: Figure 1 A flowchart illustrating a method for detecting faults in high-voltage circuit breakers in wind farms, provided by an embodiment of the present invention; Figure 2 This is a structural block diagram of a high-voltage circuit breaker fault detection system for wind farms, provided as an embodiment of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0015] Example 1:

[0016] This embodiment provides a fault detection method for high-voltage circuit breakers in wind farms, which can adapt to the complex dynamic operating conditions of wind farms. It achieves fault interference filtering, accurate anomaly identification, coupling feature decoupling, and fault trend prediction, effectively solving the technical defects of traditional detection methods such as high false alarm rate, inaccurate fault location, and inability to predict degradation trends. Figure 1 As shown, the whole process includes the following steps performed in sequence: Step S1: Acquire multi-source raw sensing data of the high-voltage circuit breaker and operating condition data reflecting the operating status of the wind farm; construct a dynamic operating condition reference domain based on the operating condition data, and compare the multi-source raw sensing data with the dynamic operating condition reference domain to filter out interference signals caused by operating condition disturbances, and output a valid abnormal signal when it is determined that there is a continuous characteristic offset.

[0017] Specifically, the multi-source raw sensing data collected in this step consists of real-time, multi-dimensional status data acquired during the operation of the high-voltage circuit breaker. This includes core sensing data such as electrical operation data, mechanical vibration data, and insulation status data. The data acquisition frequency matches the fluctuating characteristics of wind farm operating conditions, accurately capturing real-time changes in the circuit breaker's operating status. The collected operating condition data is synchronized in real-time with the wind farm's central control system operating parameters, comprehensively reflecting the wind farm's real-time operating conditions and avoiding benchmark judgment biases caused by single operating condition parameters.

[0018] Meanwhile, the dynamic operating condition reference domain abandons the traditional fixed threshold judgment mode. It is dynamically iteratively generated based on real-time updated operating condition data, adapting to instantaneous fluctuations and periodic changes in wind speed and turbine output, and accurately defining the reasonable fluctuation range of circuit breaker sensing data under different normal operating conditions. By comparing the real-time collected multi-source raw sensing data with the dynamic operating condition reference domain frame by frame and dimension by dimension, it can accurately distinguish between instantaneous signal fluctuations caused by operating condition disturbances and signal offsets caused by equipment malfunctions. Only when the signal offset state persists for a preset duration and meets the continuous offset judgment condition is the instantaneous interference signal eliminated, and a valid abnormal signal is output, significantly reducing the false alarm rate of fault detection from the source.

[0019] Step S2: In response to receiving the valid abnormal signal, extract its multi-dimensional fault features and make a preliminary judgment on the abnormality category based on the basic fault feature library.

[0020] Specifically, upon receiving a valid abnormal signal from S1, a multi-dimensional fault feature extraction program is immediately initiated. This program comprehensively extracts multi-dimensional parameters corresponding to the valid abnormal signal, including time-domain features, frequency-domain features, amplitude features, and rate of change features, ensuring complete retention of the abnormal data's feature information and preventing judgment errors caused by feature omissions. The basic fault feature library is a pre-trained, standardized database that stores standard fault feature parameters, feature forms, and corresponding abnormal category labels for various common faults and potential hazards in high-voltage circuit breakers, covering all types of faults, including electrical faults, mechanical faults, and insulation degradation faults. By matching and comparing the real-time extracted multi-dimensional fault features with the standard features in the basic fault feature library, a rapid and preliminary classification of the current equipment's abnormality category is achieved, pinpointing the fault category to which the abnormality belongs, laying the foundation for subsequent precise decoupling analysis.

[0021] Step S3: In response to the preliminary judgment result that there is a stable anomaly, the preset fault correlation matrix is ​​invoked to perform coupling and decoupling analysis on the multi-dimensional fault features corresponding to the stable anomaly, so as to separate the mutually superimposed independent fault features.

[0022] Specifically, after S2 completes the initial anomaly assessment, the assessment results undergo stability verification to eliminate false anomalies caused by random fluctuations. Deep decoupling analysis is initiated only for persistent and stable anomalies. The fault correlation matrix is ​​a quantized matrix trained on a large number of wind farm circuit breaker fault samples. It clarifies the coupling relationships, superposition weights, and interference coefficients between electrical, mechanical vibration, and insulation characteristics, accurately quantifying the mutual influence of various characteristics. Based on the fault correlation matrix, multi-dimensional coupled fault features are decomposed and analyzed, breaking the state of mutual superposition and interference among multiple fault features. This accurately isolates independent, non-interfering single fault features, solving the problems of traditional techniques failing to distinguish coupled faults and having ambiguous fault source locations.

[0023] Step S4: Based on the independent fault characteristics and the historical state time series data of the high-voltage circuit breaker, deduce the fault deterioration propagation path and predict the fault evolution trend.

[0024] Specifically, the historical status time-series data of high-voltage circuit breakers contains the operating status, anomaly records, fault data, and operating condition adaptation records throughout the equipment's entire lifecycle, comprehensively reflecting the long-term operational status change patterns. Based on the precise independent fault characteristics obtained through decoupling, and combined with similar fault evolution cases and degradation patterns in historical time-series data, the entire process of fault initiation and development is traced backward, clarifying the transmission link from initial hidden dangers to manifest faults, and identifying key nodes and inducing factors in fault degradation. Simultaneously, based on time-series evolution pattern modeling and analysis, the subsequent degradation speed, development trend, and fault risk level are predicted, achieving a technological upgrade from "post-event alarm" to "pre-event prediction," providing precise data support for predictive maintenance and early intervention and remediation of equipment.

[0025] Furthermore, the operating condition data includes at least the real-time output of the wind turbine and the instantaneous wind speed of the wind field; the dynamic operating condition reference domain is a model established based on the real-time output of the wind turbine and the instantaneous wind speed of the wind field, which characterizes the dynamic change range of each sensing data of the high-voltage circuit breaker under normal operating conditions.

[0026] Understandably, real-time wind turbine output and instantaneous wind speed are core operating conditions affecting the operation of circuit breakers in wind farms. Fluctuations in turbine output directly alter the electrical load state of circuit breakers, while sudden changes in wind speed trigger instantaneous changes in the electric field environment and equipment vibration. Both comprehensively cover the core dynamic operating conditions of wind farms. A dynamic operating condition benchmark domain constructed based on these two core parameters can accurately match special operating scenarios in wind farms and update the upper and lower limits of dynamic fluctuations in various sensing data such as electrical, mechanical, and insulation parameters of circuit breakers under normal operating conditions in real time. Compared to fixed threshold models, it perfectly adapts to the intermittent and fluctuating operating characteristics of wind farms, ensuring the accuracy of interference filtering.

[0027] Furthermore, after step S2, the method further includes: transmitting the preliminary anomaly category result back through the first negative feedback channel, and updating and correcting the dynamic operating condition reference domain based on the anomaly category result.

[0028] Specifically, an independent first negative feedback channel is set up to achieve closed-loop linkage optimization between the anomaly judgment result and the operating condition baseline domain model. After completing the initial anomaly category judgment and confirming the existence of a real anomaly in the equipment, the operating environment, signal characteristics, and anomaly category data corresponding to this anomaly are transmitted back to the training module of the dynamic operating condition baseline domain in real time. By continuously accumulating real anomaly samples, the fluctuation threshold and judgment rules of the dynamic operating condition baseline domain are iteratively corrected, and the distinction boundary between operating condition disturbances and real anomalies is optimized. This allows the baseline domain model to continuously adapt to long-term changes such as equipment aging and operating condition iteration, and continuously improve the accuracy of interference filtering and anomaly identification.

[0029] Furthermore, in step S1, when outputting a valid abnormal signal, a first-level confidence score is assigned to the valid abnormal signal.

[0030] Specifically, the first-level confidence score is used to quantify the true reliability of the valid abnormal signals output in step S1. The score is calculated based on multiple dimensions of parameters, including signal offset amplitude, offset duration, and intensity of operational disturbances. Different score ranges correspond to different levels of anomaly confidence. By setting the first-level confidence score, the filtered valid abnormal signals can be quantified in a hierarchical manner, providing a precise quantitative basis for the weighted correction of subsequent anomaly judgments and avoiding result bias caused by judging a single signal.

[0031] Furthermore, in step S2, the preliminary determination of the anomaly category based on the basic fault feature library specifically includes: weighting and correcting the preliminary determination result by combining the first-level confidence score.

[0032] Specifically, after completing the initial matching of fault features with the basic fault feature library, the first-level confidence score output by S1 is introduced as a weighting coefficient to adjust the anomaly category, fault probability, and feature matching degree obtained from the initial matching. For valid anomaly signals with high confidence, the weight of the corresponding feature matching result is increased; for suspected anomaly signals with low confidence, the matching weight is weakened while retaining space for verification, effectively improving the accuracy of the initial anomaly judgment results and reducing the probability of missed or false judgments.

[0033] In step S2, after completing the preliminary determination of the anomaly category, a secondary confidence score is assigned to the preliminary determination result.

[0034] Specifically, the secondary confidence score is used to quantify the credibility of the preliminary anomaly category determination. The score is calculated comprehensively based on parameters such as feature matching similarity, primary confidence benchmark, and operating condition adaptability, directly reflecting the accuracy level of the current anomaly category determination. As a core control parameter for subsequent coupling and decoupling analysis, the secondary confidence score enables quantitative control of the entire detection process, making fault analysis more accurate and traceable.

[0035] Furthermore, in step S3, the step of calling the preset fault correlation matrix for coupling and decoupling analysis specifically includes: adjusting the feature separation threshold in the coupling and decoupling analysis process based on the secondary confidence score.

[0036] Specifically, the feature separation threshold directly determines the accuracy of stripping multidimensional coupled fault features. The separation threshold is dynamically and adaptively adjusted according to the level of the secondary confidence score: when the secondary confidence score is high and the credibility of the anomaly judgment result is high, the feature separation threshold is tightened to improve the fine decoupling accuracy and accurately identify subtle fault feature differences; when the secondary confidence score is low and there is uncertainty in the anomaly, the feature separation threshold is relaxed to expand the feature extraction range, avoid missing subtle fault features, and take into account both the accuracy and comprehensiveness of fault detection.

[0037] In step S3, the coupling and decoupling analysis includes peeling off multi-dimensional fault features step by step according to a preset progressive order. The preset progressive order is: first, separate electrical features, then peel off mechanical vibration features, and finally extract insulation degradation features.

[0038] Specifically, based on the occurrence patterns and characteristic coupling properties of high-voltage circuit breaker faults, a scientific progressive stripping sequence is established: electrical signals have the fastest response speed and the strongest sensitivity to operating conditions, making them the most susceptible to operating condition interference and the most likely to show abnormalities first. Therefore, the separation and identification of electrical features are prioritized. Mechanical vibration features are mostly caused by equipment structural wear and movement jamming, and are often superimposed with secondary interference from electrical abnormalities. They can be accurately extracted after stripping electrical features. Insulation degradation features change slowly and are highly concealed, easily masked by electrical and mechanical signals. They are extracted last, which can completely avoid interference from the superposition of multiple features and achieve accurate and step-by-step decoupling of the three types of core fault features.

[0039] In step S3, after completing the coupling and decoupling analysis and separating the independent fault features, a three-level confidence score is assigned to the independent fault features.

[0040] Specifically, the three-level confidence score is used to quantify the authenticity and effectiveness of each independent fault feature after decoupling. It is comprehensively assigned based on parameters such as feature separation accuracy, the second-level confidence benchmark, and the fault correlation matrix matching degree, accurately characterizing the credibility of each type of independent fault feature. This score provides a quantitative weighting basis for subsequent fault tracing, degradation path deduction, and trend prediction, ensuring the accuracy of fault evolution analysis.

[0041] In step S4, the deduction of the fault degradation propagation path specifically includes: matching and tracing the evolution path of the fault from historical state time series data based on the three-level confidence scores and the independent fault characteristics.

[0042] Specifically, using high-confidence independent fault characteristics as the core search criteria, and combining three levels of confidence scores to set matching thresholds, similar fault cases are accurately matched in the historical state time-series data of the entire equipment lifecycle. By tracing the changes in fault characteristics from non-existent to present and from weak to strong in historical data, the initial causes of fault initiation, intermediate degradation stages, and explicit fault outbreak nodes are identified, fully reconstructing the complete path of fault degradation propagation and accurately locating the root cause of the fault.

[0043] Furthermore, after step S4, the method further includes: transmitting the deduced fault deterioration propagation path and the predicted fault evolution trend back through a global feedback channel to iteratively update the basic fault feature library and the fault correlation matrix.

[0044] Specifically, a global feedback channel is established to form a closed-loop optimization mechanism throughout the entire process. The independent fault features, fault degradation propagation paths, fault evolution trends, and final fault determination results obtained from this detection are used as new sample data and transmitted back in real time to the training system for the basic fault feature library and the fault correlation matrix. By continuously accumulating real-world wind farm fault cases, the sample dimensions of the fault feature library are constantly enriched, and the coupling weight parameters of the fault correlation matrix are optimized. This continuously improves the adaptability and accuracy of overall fault detection, decoupling, and prediction, enabling the method to have the ability for continuous iterative optimization.

[0045] Example 2:

[0046] This embodiment provides a high-voltage circuit breaker fault detection system for wind farms, used to implement the high-voltage circuit breaker fault detection method provided in Embodiment 1 or any possible implementation of Embodiment 1. It is adaptable to the complex and harsh operating conditions of wind farms, achieving intelligent detection throughout the entire process, including automatic filtering of operating interference, accurate anomaly judgment, decoupling of coupled faults, fault tracing, and trend prediction. This effectively improves the accuracy and timeliness of high-voltage circuit breaker fault detection, providing technical support for the safe operation and maintenance of wind farm equipment. The system adopts a modular architecture design, with each module operating independently yet interconnected and collaboratively linked. Figure 2 As shown, it specifically includes a working condition disturbance filtering module, a manifest anomaly initial judgment module, a coupled fault decoupling analysis module, and a fault source tracing and evolution prediction module. The specific functions of each module are as follows: The operating condition disturbance filtering module is used to execute step S1 in Example 1. Specifically, it is responsible for real-time acquisition of multi-source raw sensing data from the high-voltage circuit breaker and wind farm operating condition data. Based on core operating condition parameters such as real-time wind turbine output and instantaneous wind speed, it dynamically constructs and updates a dynamic operating condition reference domain, completing the comparison and analysis between the real-time sensing data and the dynamic reference domain. It accurately identifies and filters out false interference signals caused by operating condition disturbances such as sudden wind speed changes and power fluctuations, outputting only valid abnormal signals with a first-level confidence score when a persistent characteristic offset is detected. This avoids false alarms caused by operating condition fluctuations at the source, providing clean and valid data support for subsequent anomaly detection.

[0047] The explicit anomaly preliminary judgment module is communicatively connected to the operating condition disturbance filtering module and is used to execute step S2 in Embodiment 1. This module receives valid anomaly signals output by the operating condition disturbance filtering module in real time, completes comprehensive extraction of multi-dimensional fault features, and performs preliminary matching and judgment of anomaly categories based on the built-in basic fault feature library. At the same time, it combines the first-level confidence score of the valid anomaly signal to perform weighted correction on the judgment result and assigns a second-level confidence score to the preliminary judgment result to complete the stability verification of the abnormal state. Simultaneously, through the built-in first negative feedback channel, the anomaly category judgment result is sent back to the operating condition disturbance filtering module to realize iterative correction of the dynamic operating condition reference domain and form a local closed-loop optimization mechanism.

[0048] The coupled fault decoupling analysis module, communicatively connected to the explicit anomaly preliminary judgment module, is used to execute step S3 in Embodiment 1. This module receives the stable anomaly result and corresponding secondary confidence score output by the explicit anomaly preliminary judgment module, calls the system's preset fault correlation matrix, and dynamically adjusts the feature separation threshold based on the secondary confidence score. Strictly following the preset progressive order of "prioritizing the separation of electrical features, then stripping away mechanical vibration features, and finally extracting insulation degradation features," it performs step-by-step stripping and decoupling analysis on multidimensional coupled fault features, thoroughly separating mutually superimposed and interfering independent fault features, and assigning a tertiary confidence score to each independent fault feature, thus completing the accurate decomposition and quantitative identification of complex coupled faults.

[0049] The fault tracing and evolution prediction module is communicatively connected to the coupled fault decoupling and analysis module, and is used to execute step S4 in Embodiment 1. This module receives the decoupled independent fault characteristics and their corresponding three-level confidence scores, retrieves the historical state time-series data of the high-voltage circuit breaker throughout its entire lifecycle, accurately matches similar fault evolution cases, traces back and deduces the complete path of fault degradation propagation, and clarifies the root cause of the fault and the law of degradation development. Simultaneously, based on the time-series evolution model, it predicts the subsequent evolution trend, degradation rate, and risk level of the fault, outputting accurate fault prediction results. Furthermore, this module transmits the fault degradation propagation path and evolution trend data back through a global feedback channel, iteratively updating the basic fault feature library and fault correlation matrix, achieving continuous optimization and upgrading of the overall system algorithm.

[0050] Example 3:

[0051] This embodiment provides an electronic device applicable to high-voltage circuit breaker fault detection scenarios in wind farms. It can efficiently execute the high-voltage circuit breaker fault detection method for wind farms provided in Embodiment 1 or any possible implementation thereof, achieving intelligent, precise, and predictive fault detection of high-voltage circuit breakers in wind farms. Specifically, the electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it can completely implement all the detection steps and optimization logic disclosed in Embodiment 1.

[0052] Among them, the memory serves as the core unit for equipment data storage. It is used to persistently store historical data of multi-source sensing of high-voltage circuit breakers, historical data of wind farm operating conditions, basic fault feature library, fault correlation matrix, dynamic operating condition benchmark domain model parameters, various confidence score algorithm models and equipment operation log data. At the same time, it caches real-time collected dynamic operating condition data, raw sensing data and real-time detection intermediate results, providing complete data storage support for fault detection, data comparison, model iteration and trend inference.

[0053] As the core unit for equipment operation and control, the processor establishes a data interaction connection with the memory. It can call the computer programs stored in the memory to automatically complete the entire process of operation, including data acquisition, construction of dynamic operating condition reference domain, interference signal filtering, output of effective abnormal signals, extraction of multi-dimensional fault features, preliminary judgment of abnormality categories, weighted correction of confidence scores, decoupling analysis of coupled faults, fault path tracing, evolution trend prediction, and dual-channel feedback iteration. The processor has high-speed data processing capabilities, can adapt to the high-frequency dynamic changes in wind farm operating condition data, and can complete fault detection calculations in real time, ensuring the real-time and accurate nature of detection results. It can quickly identify hidden dangers and obvious faults in equipment and output reliable detection and prediction results.

[0054] This electronic device can be deployed independently in the central control room of a wind farm or integrated into the online monitoring terminal of a high-voltage circuit breaker. It is suitable for the harsh operating environments of wind farms in the Gobi Desert, mountainous areas, and offshore areas. It has the advantages of strong stability, high calculation accuracy, and adaptive iteration, and can meet the long-term online fault detection requirements of high-voltage circuit breakers in wind farms.

[0055] Example 4:

[0056] This embodiment provides a computer-readable storage medium, which is a non-volatile readable storage medium that can be widely used in various wind farm detection terminals, industrial control equipment, servers and intelligent monitoring equipment. The medium stores a computer program, which, when executed by the device processor, can completely implement all the steps and technical logic of the wind farm high-voltage circuit breaker fault detection method provided in Embodiment 1 or any possible implementation of Embodiment 1.

[0057] Specifically, this computer-readable storage medium can permanently store complete fault detection algorithm programs, dynamic baseline domain construction algorithms, confidence scoring algorithms, fault feature decoupling algorithms, time-series tracing and deduction algorithms, and feedback iterative optimization programs. It can also store core model data such as basic fault feature libraries and fault correlation matrices. The storage medium supports real-time data reading and writing, iterative updates, and long-term persistent storage, adapting to the storage needs of wind farm industrial operation scenarios. It features anti-interference, high stability, and fast read / write speeds.

[0058] When various intelligent detection devices equipped with this storage medium are powered on, the processor can quickly read and execute the computer program within the storage medium, automatically completing tasks such as filtering disturbances under the operating conditions of high-voltage circuit breakers in wind farms, identifying abnormal signals, initially judging fault characteristics, decoupling coupling characteristics, tracing the source of fault degradation, and predicting trends. Simultaneously, it achieves closed-loop iterative optimization of the model and database, completing the entire intelligent fault detection process without manual intervention. Through this computer-readable storage medium, the fault detection method of this application can be quickly ported and adapted to various industrial testing hardware devices, significantly improving the versatility, portability, and applicability of the fault detection technology for high-voltage circuit breakers in wind farms.

[0059] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for fault detection of high-voltage circuit breakers in wind farms, characterized in that, The following steps are performed in sequence: Step S1: Acquire multi-source raw sensing data of the high-voltage circuit breaker and operating condition data reflecting the operating status of the wind farm; A dynamic operating condition reference domain is constructed based on the operating condition data, and the multi-source raw sensing data is compared with the dynamic operating condition reference domain to filter out interference signals caused by operating condition disturbances, and when a persistent feature offset is determined, an effective abnormal signal is output. Step S2: In response to receiving the valid abnormal signal, extract its multi-dimensional fault features and make a preliminary judgment on the abnormality category based on the basic fault feature library; Step S3: In response to the preliminary judgment result that there is a stable anomaly, the preset fault correlation matrix is ​​invoked to perform coupling and decoupling analysis on the multi-dimensional fault features corresponding to the stable anomaly, so as to separate the mutually superimposed independent fault features. Step S4: Based on the independent fault characteristics and the historical state time series data of the high-voltage circuit breaker, deduce the fault deterioration propagation path and predict the fault evolution trend.

2. The method according to claim 1, characterized in that, The operating condition data includes at least the real-time output of the wind turbine and the instantaneous wind speed of the wind field; the dynamic operating condition reference domain is a model established based on the real-time output of the wind turbine and the instantaneous wind speed of the wind field, which characterizes the dynamic change range of each sensing data of the high-voltage circuit breaker under normal operating conditions.

3. The method according to claim 2, characterized in that, Following step S2, the following is also included: The preliminary anomaly category result is transmitted back through the first negative feedback channel, and the dynamic operating condition reference domain is updated and corrected based on the anomaly category result.

4. The method according to claim 3, characterized in that, In step S1, when outputting a valid abnormal signal, a first-level confidence score is also assigned to the valid abnormal signal. In step S2, the preliminary determination of the anomaly category based on the basic fault feature library specifically includes: weighting and correcting the preliminary determination result by combining the first-level confidence score.

5. The method according to claim 4, characterized in that, In step S2, after completing the preliminary determination of the anomaly category, a secondary confidence score is assigned to the preliminary determination result; In step S3, the step of calling the preset fault correlation matrix for coupling and decoupling analysis specifically includes: adjusting the feature separation threshold in the coupling and decoupling analysis process based on the secondary confidence score.

6. The method according to claim 5, characterized in that, In step S3, the coupling and decoupling analysis includes peeling off multi-dimensional fault features step by step according to a preset progressive order. The preset progressive order is: first, separate electrical features, then peel off mechanical vibration features, and finally extract insulation degradation features.

7. The method according to claim 6, characterized in that, In step S3, after completing the coupling and decoupling analysis and separating the independent fault features, a three-level confidence score is assigned to the independent fault features. In step S4, the deduction of the fault degradation propagation path specifically includes: matching and tracing the evolution path of the fault from historical state time series data based on the three-level confidence scores and the independent fault characteristics.

8. The method according to claim 7, characterized in that, Following step S4, the following is also included: The deduced fault degradation propagation path and predicted fault evolution trend are fed back through the global feedback channel to iteratively update the basic fault feature library and the fault correlation matrix.

9. A fault detection system for high-voltage circuit breakers in wind farms, characterized in that, For implementing the method as described in any one of claims 1 to 8, comprising: The operating condition disturbance filtering module is used to perform step S1; The explicit anomaly initial judgment module is communicatively connected to the operating condition disturbance filtering module and is used to execute step S2; The coupling fault decoupling analysis module is communicatively connected to the explicit anomaly preliminary judgment module and is used to execute step S3; The fault tracing and evolution prediction module is communicatively connected to the coupled fault decoupling analysis module and is used to execute step S4.

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

Images