A fault prediction method and device for reducing the malfunction probability of a digitized gas relay and a medium

Abstract

The present application relates to the field of transformer relay protection, and particularly relates to a fault prediction method and device for reducing the false action probability of a digitized gas relay and a medium. The present application first collects data and then performs cleaning processing, uses multiple filtering methods to eliminate interference, analyzes the filtered data based on a 3 sigma rule and an isolated forest algorithm, identifies abnormal data points, eliminates the abnormal data points, divides the signal data into a high confidence area and a low confidence area based on a preset fault determination threshold, directly generates a preliminary fault determination result for the signal data in the high confidence area based on a rule determination method, inputs the signal data in the low confidence area into a deep learning classification model, analyzes the low confidence signal data using the pre-trained deep learning classification model, and outputs a fault classification result. The present application reduces the false alarm probability by processing the detection data.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of transformer relay protection, and in particular to a fault prediction method and device for reducing the misoperation probability of a digital gas relay for a transformer, and a medium. BACKGROUND

[0002] With the rapid development of digitalization and intelligentization of power systems, the reliability and accuracy of gas relays, which are important protection devices for transformers and other oil-immersed power equipment, directly affect the safe operation of the equipment. Gas relays mainly monitor the gas release rate and related signals generated during equipment operation to determine whether there is a potential fault. However, the misoperation rate of digital gas relays in the prior art is relatively high, which seriously affects the operational efficiency and safety of power equipment, especially in complex working conditions.

[0003] Currently, most gas relays use simple signal analysis and fixed threshold determination methods to determine whether to trigger a protection action. The traditional method can effectively respond to some common faults, but has significant limitations. First, the existing technology has a simple processing method for raw signal data, which is difficult to deal with data interference in complex environments, such as high-frequency noise, nonlinear interference, and signal fluctuations, which can lead to misjudgment. The misoperation rate of gas relays significantly increases in high-temperature, high-humidity, or strong electromagnetic interference complex conditions. Second, the existing method lacks the ability to identify abnormal data, and cannot accurately eliminate noise and abnormal points in the sampling data, which can lead to inaccuracy in protection actions due to data quality problems. In recent years, some research has introduced data processing and intelligent algorithms to analyze gas relay signals, trying to improve the accuracy of signal determination through machine learning and statistical methods. However, the following problems have been exposed in practical applications: On the one hand, most methods fail to fully combine the characteristics of high and low confidence regions, making it difficult to effectively balance the real-time and accuracy of data in various complex environments. On the other hand, the existing technology lacks a systematic data preprocessing strategy, which fails to optimize the entire process from data collection, cleaning to determination. In addition, the current intelligent method is inefficient in processing large-scale real-time data, and cannot quickly adapt to dynamic changes in different working conditions, leading to delayed or missed protection actions.

[0004] In summary, the existing technology has significant shortcomings in data processing accuracy, signal analysis intelligence level, and real-time and reliability of protection actions in complex working conditions, which seriously affects the practical application effect of digital gas relays. A systematic and efficient solution is needed to improve the reliability and accuracy of digital gas relays in complex working conditions. SUMMARY

[0005] In view of the defects of the prior art, the application provides a fault prediction method and device for reducing the malfunction probability of a digitized gas relay and a medium, which reduces the false alarm probability and improves the fault prediction accuracy by processing detection data.

[0006] To solve the technical problem, the application adopts the technical scheme of a fault prediction method for reducing the malfunction probability of a digitized gas relay, comprising the following steps:

[0007] S1, collecting real-time signal data of equipment operation, including gas release rate data, temperature data, humidity data and electromagnetic signal strength data, through a sensor module of the digitized gas relay, to form a raw signal data set containing multiple dimensions;

[0008] S2, performing cleaning processing on the collected raw signal data set, using an adaptive filtering method to suppress high-frequency noise, using a signal smoothing algorithm to eliminate low-amplitude fluctuations, and reducing nonlinear interference in the signal through a segmented processing method;

[0009] S3, analyzing the raw signal data set after data cleaning based on the 3σ rule and the isolated forest algorithm, identifying abnormal data points in the raw signal data set, eliminating the abnormal data points from the raw signal data set, and generating a raw signal data set after abnormality elimination;

[0010] S4, analyzing the raw signal data set after abnormality elimination through a rule determination method, dividing the signal data into a high-confidence region and a low-confidence region based on a preset fault determination threshold, the signal data in the high-confidence region being normal operation state data conforming to the rule, and the signal data in the low-confidence region being data difficult to classify;

[0011] S5, for the signal data in the high-confidence region, directly generating a preliminary fault determination result based on the rule determination method, and for the signal data in the low-confidence region, inputting the signal data into a deep learning classification model, analyzing the low-confidence signal data by using a pre-trained deep learning classification model, and outputting a fault classification result;

[0012] S6, fusing the preliminary fault determination result and the fault classification result to generate a final fault determination result;

[0013] S7, recording the accuracy of the determination according to the final fault determination result and the actual operation state of the gas relay, dynamically adjusting the threshold parameters of the rule determination method and the classification weights of the deep learning classification model, and continuously optimizing the high-confidence region identification and grading determination strategy;

[0014] S8, transmitting the final fault determination result to the control module of the gas relay as a basis for triggering a protection action.

[0015] Further, step S1 is specifically:

[0016] S11, collect signal data X through the sensor module of the digital gas relay, respectively acquire gas release rate data, temperature data, humidity data and electromagnetic signal intensity data, and record the sampling time of each signal data ,

[0017] ,

[0018] Wherein represents the value of the ith signal at the jth sampling point, m is the signal type, and n is the number of sampling points;

[0019] S12, normalize the signal data X to generate a normalized signal matrix ;

[0020] S13, according to the time stamp of the real-time signal data segment the normalized signal matrix by time sequence, each segment includes data of k consecutive sampling points, forming a segmented matrix set :

[0021] ,

[0022] Wherein q is the segment index, represents the index of the sampling point contained in the segment, represents the total number of segmented signal data;

[0023] S14, store p segmented matrices as an original signal data set D:

[0024] .

[0025] Further, step S2 is specifically:

[0026] S21, use an adaptive filtering method to suppress high-frequency noise in the original signal data set, to obtain a filtered signal matrix ;

[0027] S22, perform signal smoothing processing on the filtered signal matrix to obtain a smoothed signal matrix :

[0028] ,

[0029] Wherein is the signal value of the ith row and jth column of the matrix , To smooth the window width, The signal value within the window;

[0030] S23. Signal matrix after signal smoothing. Nonlinear interference detection is performed by calculating the nonlinear interference coefficient. To determine the matrix Are there significant nonlinear disturbances?

[0031] ,

[0032] in For matrix The mean of the i-th row, where m is the number of signal types and k is the number of columns in the matrix. Exceeding the preset threshold If so, it is considered that nonlinear interference exists;

[0033] S24. For signal matrices with nonlinear interference The segmented reconstruction algorithm is applied to generate a signal matrix with nonlinear interference removed. For a matrix without nonlinear interference, directly let ;

[0034] S25, All interference-removed signal matrices Reassembled into a cleaned signal data set .

[0035] Furthermore, step S3 specifically includes:

[0036] S31. The signal data set after cleaning and processing Each signal matrix Perform statistical analysis and calculate the mean of each signal. and standard deviation ;

[0037] S32. Based on the 3σ rule, for each signal value in the signal matrix Perform preliminary anomaly detection to determine if the conditions are met:

[0038] ;

[0039] If the above conditions are met, then the signal value Mark as an outlier data point;

[0040] S33, the signal matrix after preliminary anomaly detection Further anomaly detection is performed using the Isolation Forest algorithm, constructing an Isolation Forest (IF) model, and then processing the signal matrix. Input the Isolation Forest model IF, and calculate the anomaly score of each sampling point :

[0041] ,

[0042] wherein is the multi-dimensional anomaly score of the jth sampling point, T is the set of decision trees in the Isolation Forest, are any two nodes in the decision tree, , is the signal matrix is the sampling value of the sample corresponding to the internal node t and t' at the jth sampling point, is the Euclidean distance of the sample corresponding to the nodes t and t' at the jth sampling point, is the difference in path depth of the nodes t and t' in the Isolation Forest model, which is used to measure the degree of isolation between samples, is a positive number;

[0043] If , wherein is the anomaly score threshold, all signal values of the corresponding sampling point j are marked as abnormal data points;

[0044] S34, all abnormal data points detected by the 3σ rule and the Isolation Forest algorithm are removed from the signal matrix to generate an anomaly-removed signal matrix ;

[0045] S35, all anomaly-removed signal matrices are recombined to form an anomaly-removed original signal data set .

[0046] Further, step S4 is specifically:

[0047] S41, rule-based analysis is performed on each signal matrix in the anomaly-removed original signal data set , and whether each signal value satisfies the high-confidence condition based on a preset fault judgment threshold :

[0048] ;

[0049] wherein is the mean value of the ith signal, is the preset fault judgment threshold for the ith signal; if the above condition is satisfied, the signal value is marked as a high-confidence signal value and recorded in the high-confidence region set ; otherwise, mark as low confidence signal value, record into low confidence region set

[0050] S42, form high confidence region matrix and low confidence region matrix for all signal values of signal matrix

[0051]

[0052] wherein represents high confidence region signal values in signal matrix that meet rule determination, wherein represents low confidence region signal values in signal matrix that do not meet rule determination;

[0053] S43, store high confidence region matrix and low confidence region matrix of each signal matrix according to matrix index q respectively, form high confidence region signal data set and low confidence region signal data set

[0054]

[0055] wherein p is total number of signal matrix segments.

[0056] Further, step S5 is specifically:

[0057] S51, for each high confidence region matrix in high confidence region signal data set, calculate preliminary fault determination result based on rule determination method:

[0058]

[0059] wherein represents signal value set that meets rule determination; if, determine that the high confidence region is in normal operation state, otherwise determine as potential fault state,

[0060] is high confidence region normal state determination proportion threshold; S52, for each low confidence region matrix in low confidence region signal data set, calculate preliminary fault determination result based on rule determination method:

[0061] ​​​​​​​​​​​​​​​​​​​​An input deep learning classification model is used to classify the low-confidence region matrix, and a fault classification result is generated .

[0062] Further, step S6 is specifically:

[0063] S61, obtaining a preliminary fault determination result set of the high-confidence region and a fault classification result set of the low-confidence region , respectively calculating the comprehensive determination result weight of the high-confidence region and the comprehensive classification result weight of the low-confidence region :

[0064] ,

[0065] wherein, is the number of high-confidence region signal values, is the number of low-confidence region signal values;

[0066] S62, according to the comprehensive determination result weight and the comprehensive classification result weight of the low-confidence region , the preliminary fault determination result set of the high-confidence region and the fault classification result set of the low-confidence region are weighted and fused to generate a comprehensive fault score :

[0067] ;

[0068] wherein, is the fault score of the high-confidence region, which is directly calculated based on the rule determination result, if is a potential fault, then , otherwise ; is the fault classification probability distribution output of the low-confidence region;

[0069] S63, setting a comprehensive fault score threshold judging the comprehensive fault score of each segmented matrix q , if , it is considered that the segmented fault state corresponds to the final fault determination result set :

[0070] ,

[0071] wherein is a dynamic adjustment coefficient for comprehensively evaluating the global influence of the segment,

[0072] S64. Record the individual judgment results for high-confidence regions and low-confidence regions, and combine them with the comprehensive fault judgment result set. Output a final report, including the operating status of each segment, potential fault locations, and fault category information.

[0073] Furthermore, step S8 specifically includes:

[0074] S81. Obtain the comprehensive fault judgment result set. The state information corresponding to each segment index q in the set is encoded to generate a fault state encoding matrix E:

[0075] ,

[0076] in, This represents the fault status code for segment q, where 1 indicates a fault status and 0 indicates a normal status.

[0077] S82. Combine the fault state coding matrix E with the segment matrix index q and the corresponding timestamp. Correlate and generate a fault trigger signal sequence ;

[0078] S83, Fault trigger signal sequence Input the gas relay control module and analyze its fault status information. and corresponding time To trigger the decision to take protective action; when When this happens, the control module immediately executes protective actions, including power-off operation, alarm triggering, or data storage;

[0079] S84. Set up the comprehensive fault judgment results and fault trigger signal sequence The parsing results are stored as historical execution records;

[0080] S85. Adjust the comprehensive fault score threshold based on the operational feedback from the gas relay control module. Or dynamic adjustment coefficient Optimize the next step of the fault diagnosis process.

[0081] The present invention also discloses a fault prediction device for reducing the probability of malfunction of digital gas relays, comprising a processor and a memory storing program instructions, wherein the processor is configured to execute the fault prediction method for reducing the probability of malfunction of digital gas relays as described above when running the program instructions.

[0082] The present invention also discloses a storage medium storing program instructions, which, when executed, perform the fault prediction method described above for reducing the probability of malfunction of digital gas relays.

[0083] The beneficial effects of the present application are:

[0084] The present application divides the signal data set after abnormality rejection into high-confidence region and low-confidence region, uses a rule determination method to process high-confidence region signals, and uses a deep learning classification model to analyze low-confidence region signals, thereby realizing hierarchical processing of data. Compared with the unified analysis of all data by a single algorithm in the prior art, the most suitable processing method can be selected according to the confidence of the data, the calculation complexity is reduced while ensuring the determination accuracy, and the determination efficiency is improved. In complex working conditions, the misoperation rate caused by the uncertainty of low-confidence region signals is effectively reduced, the reliability and adaptability of the gas relay are significantly enhanced, and experiments show that the fault determination accuracy of the high-confidence region is improved by 12% after using the method, and the overall calculation time is reduced by 35%.

[0085] The present application proposes a systematic data cleaning and fault result fusion algorithm. First, high-frequency noise and nonlinear interference in the signal are eliminated through adaptive filtering and segmented smoothing. Then, abnormal data is rejected through 3σ rule and isolated forest algorithm. Finally, based on the determination results of high-confidence and low-confidence, a comprehensive fault score is generated using a weighted fusion method, and the final fault judgment result is obtained. Compared with the single abnormality detection and classification method in the prior art, the method dynamically optimizes the algorithm parameters at each step to adapt to changing working conditions, ensuring the robustness of data processing. Experimental results show that the detection accuracy of the abnormality rejection algorithm of the present application in a complex noise environment is improved by 15%, and the accuracy of the final fault classification is improved by about 10%.

[0086] The present application designs a comprehensive fault score model, introduces a dynamic adjustment coefficient and a comprehensive weight, and can dynamically optimize the scoring mechanism according to real-time data, uniformly evaluate global data and adjust the determination threshold. Compared with the fixed threshold protection determination strategy in the prior art, the method of the present application is more flexible and can maintain consistent determination standards under different operating states and data distributions, avoiding protection action errors caused by the insufficient adaptability of a single fixed threshold. Tests show that under multiple working conditions, the dynamic adjustment mechanism of the present application can reduce the misjudgment rate by 20% and the missed judgment rate by 15%, significantly improving the reliability of the protection action of the gas relay. BRIEF DESCRIPTION OF DRAWINGS

[0087] Figure 1 A flowchart of a data preprocessing method for reducing the misoperation probability of a digital gas relay according to the present application;

[0088] Figure 2 A schematic diagram of the device described in Example 2. DETAILED DESCRIPTION

[0089] The application will be further described below in conjunction with the drawings and specific embodiments.

[0090] Embodiment 1

[0091] The embodiment discloses a fault detection method for reducing the malfunction probability of a digital gas relay, as shown in the following steps: Figure 1

[0092] S1, collecting real-time signal data of equipment operation through a sensor module of the digital gas relay, including gas release rate data, temperature data, humidity data and electromagnetic signal intensity data, to form a raw signal data set containing multiple dimensions;

[0093] S2, performing cleaning processing on the collected raw signal data set, using an adaptive filtering method to suppress high-frequency noise, using a signal smoothing algorithm to eliminate low-amplitude fluctuations, and reducing nonlinear interference in the signal through a segmented processing method;

[0094] S3, analyzing the raw signal data set after data cleaning based on the 3σ rule and the Isolation Forest algorithm, identifying abnormal data points in the raw signal data set, eliminating the abnormal data points from the raw signal data set, and generating a raw signal data set after abnormality elimination;

[0095] S4, analyzing the raw signal data set after abnormality elimination through a rule determination method, dividing the signal data into a high-confidence region and a low-confidence region based on a preset fault determination threshold, the signal data in the high-confidence region being normal operation state data conforming to the rules, and the signal data in the low-confidence region being data difficult to classify;

[0096] S5, for the signal data in the high-confidence region, directly generating a preliminary fault determination result based on the rule determination method, and for the signal data in the low-confidence region, inputting the signal data into a deep learning classification model, analyzing the low-confidence signal data using a pre-trained deep learning classification model, and outputting a fault classification result;

[0097] S6, fusing the preliminary fault determination result and the fault classification result to generate a final fault determination result;

[0098] S7, recording the accuracy of the determination according to the final fault determination result and the actual operation state of the gas relay, dynamically adjusting the threshold parameters of the rule determination method and the classification weights of the deep learning classification model, and continuously optimizing the high-confidence region identification and grading determination strategy;

[0099] S8, transmitting the final fault determination result to a control module of the gas relay as a basis for triggering a protection action. ​

[0100] In this embodiment, step S1 is specifically:

[0101] S11, collect signal data X through the sensor module of the digital gas relay, respectively acquire gas release rate data, temperature data, humidity data and electromagnetic signal intensity data, and record the sampling time of each signal data :

[0102] ,

[0103] Wherein represents the value of the ith signal at the jth sampling point, m is the signal type, and n is the number of sampling points;

[0104] S12, normalize the signal data X to generate a normalized signal matrix ;

[0105] S13, according to the time stamp of the real-time signal data segment the normalized signal matrix by time sequence, each segment includes data of k consecutive sampling points, forming a segmented matrix set :

[0106] ,

[0107] Wherein q is the segment index, represents the index of the sampling point contained in the segment, represents the total number of segmented signal data, represents the floor function;

[0108] S14, store p segmented matrices as an original signal data set D:

[0109] .

[0110] In this embodiment, step S2 is specifically:

[0111] S21, use an adaptive filtering method to suppress high-frequency noise in each segmented matrix in the original signal data set D, to obtain a filtered signal matrix ;

[0112] S22, perform signal smoothing processing on the filtered signal matrix to obtain a smoothed signal matrix :

[0113] ,

[0114] Wherein is the matrix the signal value in the i-th row and j-th column, is the smoothing window width, is the signal value within the window;

[0115] S23, the signal matrix after signal smoothing processing nonlinear interference detection is performed by calculating the nonlinear interference coefficient to determine whether the matrix exists significant nonlinear interference:

[0116] ,

[0117] wherein is the mean value of the i-th row of the matrix, m is the number of signal types, k is the number of matrix columns, and if exceeds a preset threshold , it is considered that there is nonlinear interference;

[0118] S24, for the signal matrix with nonlinear interference a piecewise reconstruction algorithm is applied to generate a signal matrix without nonlinear interference , for the matrix without nonlinear interference, directly ;

[0119] S25, all the signal matrices after removing interference are recombined into a signal data set after cleaning processing .

[0120] In this embodiment, step S3 is specifically:

[0121] S31, statistical analysis is performed on each signal matrix of the signal data set after cleaning processing , and the mean value and standard deviation of each signal are calculated;

[0122] S32, based on the 3σ rule, preliminary anomaly detection is performed on each signal value in the signal matrix to determine whether the condition is met:

[0123] ;

[0124] If the above condition is met, the signal value is marked as an abnormal data point;

[0125] S33, for the signal matrix after preliminary anomaly detection , an isolated forest algorithm is applied for further anomaly detection to construct an isolated forest model IF, and the signal matrix​ Input the Isolation Forest model IF, and calculate the anomaly score of each sampling point :

[0126] ,

[0127] wherein is the multi-dimensional anomaly score of the jth sampling point, T is a set of decision trees in the Isolation Forest, are any two nodes in the decision tree, , is a signal matrix , wherein is the Euclidean distance of the samples corresponding to nodes t and t' at the jth sampling point, is the difference in path depth of nodes t and t' in the Isolation Forest model, used to measure the degree of isolation between samples, is a positive number;

[0128] If , wherein is an anomaly score threshold, all signal values of the corresponding sampling point j are marked as abnormal data points;

[0129] S34, all abnormal data points detected by the 3σ rule and the Isolation Forest algorithm are removed from the signal matrix to generate an anomaly-removed signal matrix ;

[0130] S35, all anomaly-removed signal matrices are recombined to form an anomaly-removed original signal data set .

[0131] In this embodiment, step S4 is specifically:

[0132] S41, each signal matrix in the anomaly-removed original signal data set is subjected to rule determination analysis, and whether each signal value satisfies a high confidence condition is determined based on a preset fault determination threshold :

[0133] ;

[0134] wherein is the mean value of the ith signal, is the preset fault determination threshold for the ith signal; if the above condition is satisfied, the signal value is marked as high-confidence signal values and recorded in a high-confidence region set ; otherwise, marked as low-confidence signal values and recorded in a low-confidence region set ; ;

[0135] S42, for all signal values of the signal matrix , a high-confidence region matrix and a low-confidence region matrix are formed:

[0136] ,

[0137] wherein represents high-confidence region signal values in the signal matrix that meet the rule determination, represents low-confidence region signal values in the signal matrix that do not meet the rule determination;

[0138] S43, the high-confidence region matrix and the low-confidence region matrix of each signal matrix are respectively stored according to the matrix index q, forming a high-confidence region signal data set and a low-confidence region signal data set :

[0139] ;

[0140] wherein p is the total number of segments of the signal matrix.

[0141] In this embodiment, step S5 is specifically:

[0142] S51, for each high-confidence region matrix in the high-confidence region signal data set , a preliminary fault determination result is calculated based on the rule determination method:

[0143] ,

[0144] wherein represents a signal value set that meets the rule determination;

[0145] if , it is determined that the high-confidence region is in a normal operating state, otherwise it is determined to be in a potential fault state, is a high-confidence region normal state determination proportion threshold value;

[0146] S52, for each low-confidence region matrix in the low-confidence region signal data set The input deep learning classification model is used to classify the low-confidence area matrix, and a fault classification result is generated .

[0147] In this embodiment, step S6 is specifically:

[0148] S61, obtaining a preliminary fault determination result set of the high-confidence area and a fault classification result set of the low-confidence area , respectively calculating the comprehensive determination result weight of the high-confidence area and the comprehensive classification result weight of the low-confidence area :

[0149] ,

[0150] wherein, is the number of signal values of the high-confidence area, is the number of signal values of the low-confidence area;

[0151] S62, according to the comprehensive determination result weight and the comprehensive classification result weight of the low-confidence area , the preliminary fault determination result set of the high-confidence area and the fault classification result set of the low-confidence area are weighted and fused to generate a comprehensive fault score :

[0152] ;

[0153] wherein, is the fault score of the high-confidence area, which is directly calculated based on the rule determination result, if is a potential fault, then , otherwise ; is the fault classification probability distribution output of the low-confidence area;

[0154] S63, setting a comprehensive fault score threshold judging the comprehensive fault score of each segmented matrix q , if , it is considered that the segmented fault state corresponds to the fault state, and a final fault determination result set is generated :

[0155] ,

[0156] wherein is a dynamic adjustment coefficient, used for comprehensive evaluation of the global influence of the segmented,

[0157] S64, record the individual determination results of high and low confidence areas, and combine the comprehensive fault determination result set Output the final report, including the running state, potential fault location and fault category information of each segment.

[0158] In this embodiment, step S8 is specifically:

[0159] S81, obtain the comprehensive fault determination result set Encode the state information corresponding to each segment index q in the set to generate a fault state encoding matrix E:

[0160] ,

[0161] Wherein, The fault state encoding of segment q, 1 represents the fault state, and 0 represents the normal state;

[0162] S82, associate the fault state encoding matrix E with the segment matrix index q and the corresponding timestamp To generate a fault trigger signal sequence ;

[0163] S83, input the fault trigger signal sequence into the gas relay control module, analyze the fault state information therein and the corresponding time , to trigger the decision of protection action; when , the control module immediately executes the protection action, including power-off operation, alarm triggering or data storage;

[0164] S84, store the analysis results of the comprehensive fault determination result set and the fault trigger signal sequence as historical operation records;

[0165] S85, according to the running feedback of the gas relay control module, adjust the comprehensive fault score threshold or the dynamic adjustment coefficient , optimize the next step of fault determination process.

[0166] Embodiment 2

[0167] Combined with Figure 2As shown, the embodiment of the present disclosure provides a fault prediction device for reducing the probability of misoperation of a digital gas relay, which comprises a processor and a memory. Optionally, the device can further comprise a communication interface and a bus. The processor, the communication interface and the memory can communicate with each other through the bus. The communication interface can be used for information transmission. The processor can call the logic instructions in the memory to execute the fault prediction method for reducing the probability of misoperation of a digital gas relay in embodiment 1.

[0168] In addition, the logic instructions in the memory described above can be implemented in the form of a software functional unit and sold or used as an independent product, which can be stored in a computer readable storage medium.

[0169] The memory, as a computer readable storage medium, can be used to store software programs, computer executable programs, such as program instructions / modules corresponding to the method in the embodiment of the present disclosure. The processor executes the program instructions / modules stored in the memory, thereby performing functional applications and data processing, that is, implementing the fault prediction method for reducing the probability of misoperation of a digital gas relay in embodiment 1.

[0170] The memory can include a program storage area and a data storage area, wherein the program storage area can store an operating system and at least one application required by a function; the data storage area can store data created according to the use of the terminal device, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory.

[0171] Embodiment 3

[0172] The embodiment of the present disclosure provides a computer readable storage medium, which stores computer executable instructions, and the computer executable instructions are set to execute the fault prediction method for reducing the probability of misoperation of a digital gas relay in embodiment 1.

[0173] The computer readable storage medium described above can be a transitory computer readable storage medium or a non-transitory computer readable storage medium.

[0174] The technical solutions of the embodiments of the present disclosure can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes one or more instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method disclosed in the embodiments of the present disclosure. The aforementioned storage medium can be a non-transitory storage medium, including: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes, or can be a transitory storage medium.

[0175] The above description and drawings sufficiently illustrate the embodiments of the present disclosure to enable one skilled in the art to practice them. Other embodiments can include structural, logical, electrical, process, and other changes. The embodiments represent only a few of the possible variations. Individual components and functions are optional unless explicitly required, and the order of operations can be changed. Parts and features of some embodiments can be included in or replace parts and features of other embodiments. Also, the words used in this application are only used to describe the embodiments and not to limit the scope of protection. As used in the description herein, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms as well. Similarly, as used in this application, the term "and / or" refers to any and all possible combinations of one or more associated listed items. In addition, when used in this application, the term "comprise" and its variants "comprises" and / or comprises, etc. refer to the presence of the stated features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups of these. Without more limitations, the element defined by the phrase "comprising a" does not exclude the presence of additional identical elements in the process, method, or device that includes the stated element. In this document, each embodiment focuses on the differences from other embodiments, and the same or similar parts between various embodiments can be referred to each other. For the method, product, etc. disclosed in the embodiments, if it corresponds to the method part disclosed in the embodiments, the relevant part can be referred to the description of the method part.

[0176] Those skilled in the art can clearly understand that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software mode can depend on specific application and design constraints of the technical solution. The skilled person can use different methods for each specific application to realize the described functions, but such implementation should not be considered beyond the scope of the embodiments of the present disclosure. The skilled person can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, devices and units can refer to the corresponding processes in the foregoing method embodiments, which will not be described here again.

Claims

1. A fault prediction method for reducing the probability of malfunction of a digital gas relay, characterized by, The method comprises the following steps: S1, collecting real-time signal data of equipment operation through a sensor module of a digital gas relay, including gas release rate data, temperature data, humidity data and electromagnetic signal intensity data, to form a raw signal data set containing multiple dimensions; S2, cleaning the collected raw signal data set, using an adaptive filtering method to suppress high-frequency noise, using a signal smoothing algorithm to eliminate low-amplitude fluctuations, and reducing nonlinear interference in the signal through segmented processing; S3, analyzing the raw signal data set after data cleaning based on the 3σ rule and the Isolation Forest algorithm, identifying abnormal data points in the raw signal data set, removing the abnormal data points from the raw signal data set, and generating a raw signal data set after removing the abnormal data points; S4, analyzing the raw signal data set after removing the abnormal data points through a rule determination method, dividing the signal data into a high-confidence region and a low-confidence region based on a pre-set fault determination threshold, the signal data in the high-confidence region being normal operation state data conforming to the rules, and the signal data in the low-confidence region being data difficult to classify; S5, for the signal data in the high-confidence region, directly generating a preliminary fault determination result based on the rule determination method, and for the signal data in the low-confidence region, inputting it into a deep learning classification model, analyzing the low-confidence signal data using a pre-trained deep learning classification model, and outputting a fault classification result; S6, fusing the preliminary fault determination result and the fault classification result to generate a final fault determination result; S7, according to the final fault determination result and the actual operation state of the gas relay, recording the accuracy of the determination, dynamically adjusting the threshold parameters of the rule determination method and the classification weights of the deep learning classification model, and continuously optimizing the high-confidence region identification and grading determination strategy; S8, transmitting the final fault determination result to the control module of the gas relay as a basis for triggering a protection action.

2. The method of claim 1, wherein the method further comprises: Step S1 is specifically: S11, collect signal data X through the sensor module of the digital gas relay, respectively acquire gas release rate data, temperature data, humidity data and electromagnetic signal intensity data, and record the sampling time of each signal data , ， wherein denotes the value of the ith signal at the jth sample point, m is the signal class, and n is the number of sample points; S12, normalizing the signal data X to generate a normalized signal matrix ; S13, time stamp of the real-time signal data normalizing the signal matrix performing time series segmentation, each segment comprising data of k consecutive sampling points, forming a set of segment matrices : ， where q is a segment index, an index representing a sample point included in the segment, an index representing a total number of segments into which the signal data is divided; S14, store the p segmented matrices as a raw signal data set D: 。 3. The method of claim 1, wherein the method further comprises: Step S2 is specifically: S21, using an adaptive filtering method to suppress high-frequency noise in the original signal data set, obtaining a filtered signal matrix ; S22, performing signal smoothing processing on the filtered signal matrix performing signal smoothing processing, to obtain a signal matrix after smoothing processing : ， wherein is a matrix is the signal value in the ith row and jth column, is the smoothing window width, is the signal value within the window; S23, the signal matrix after smoothing processing on the signal Nonlinear interference detection is performed by calculating a nonlinear interference coefficient to determine whether the matrix exists significant nonlinear interference: ， wherein is a matrix is the mean value of the i-th row, m is the number of signal types, k is the number of matrix columns, and if exceeds a predetermined threshold then it is considered that there is a nonlinear interference; S24, the signal matrix with nonlinear interference Apply the piecewise reconstruction algorithm to generate the signal matrix without nonlinear interference , the matrix without nonlinear interference, directly let ; S25, recombining all the interference-removed signal matrices into a cleaned signal data set .

4. The method of claim 3, wherein the method further comprises: Step S3 is specifically: S31, performing statistical analysis on the signal data set after cleaning each signal matrix performing statistical analysis, calculating the mean value of each signal and standard deviation ; S32, based on the 3σ rule, each signal value in the signal matrix performing preliminary anomaly detection, determining whether the condition is met: ； If the above conditions are met, the signal value is marked as an outlier data point; S33, the signal matrix after preliminary anomaly detection Further anomaly detection is performed using the Isolation Forest algorithm, constructing an Isolation Forest (IF) model, and then processing the signal matrix. Input the Isolation Forest model IF and calculate the anomaly score for each sampling point. : ， wherein is the multi-dimensional anomaly score of the jth sample point, T is the set of decision trees in the Isolation Forest, are any two nodes in the decision tree, respectively, , is the signal matrix is the sample value of the samples corresponding to the inner nodes t and t' at the jth sample point, is the Euclidean distance of the samples corresponding to the nodes t and t' at the jth sample point, is the difference of the path depth of the nodes t and t' in the Isolation Forest model, which is used to measure the isolation degree between samples, is a positive number; If wherein is an anomaly score threshold, all signal values of the corresponding sampling point j are flagged as an anomaly data point; S34, remove all the abnormal data points detected by 3σ rule and Isolation Forest algorithm from the signal matrix to generate the signal matrix after removing the abnormal data ; S35, signal matrix after removing all anomalies recombine to form a set of original signal data after removing anomalies .

5. The method of claim 4, wherein the method further comprises: Step S4 is specifically: S41、performing rule determination analysis on the original signal data set after removing the abnormal data each signal matrix in the set performing rule determination analysis on each signal value based on a preset fault determination threshold determining whether a high confidence condition is met: ； wherein, is the mean value of the i-th signal, is a preset fault decision threshold for the i-th signal; if the above condition is satisfied, the is marked as a high-confidence signal value and recorded in a high-confidence region set ; otherwise, the is marked as a low-confidence signal value and recorded in a low-confidence region set ; S42, form a high-confidence region matrix for all signal values of the signal matrix S44, form a low-confidence region matrix for all signal values of the signal matrix S46, form a high-confidence region matrix for all signal values of the signal matrix S48, form a low-confidence region matrix for all signal values of the signal matrix ， wherein representing signal matrix high confidence region signal values in which representing signal matrix low confidence region signal values in which the rule determination is not satisfied; S43, a high-confidence region matrix for each signal matrix and a low-confidence region matrix are stored separately by matrix index q, forming a high-confidence region signal data set and a low-confidence region signal data set : ； Where p is the total number of segmented matrices of the signal matrix.

6. The method of claim 5, wherein the method further comprises: Step S5 is specifically: S51, for each high-confidence region matrix in the high-confidence region signal data set S51, for each high-confidence region matrix in the high-confidence region signal data set S51, for each high-confidence region matrix in the high-confidence region signal data set ， wherein represents a set of signal values that comply with the rule decision; If then the high confidence region is determined to be in a normal operating state, otherwise it is determined to be in a potential fault state, is a high confidence region normal state determination proportion threshold. S52, generate a low-confidence region signal data set each low-confidence region matrix in the low-confidence region signal data set input a deep learning classification model, perform fault classification on the low-confidence region matrix by using the deep learning classification model, and generate a fault classification result .

7. A fault prediction method for reducing the probability of malfunction of a digital gas relay according to claim 6, characterized in that: Step S6 is specifically: S61、obtain a preliminary fault determination result set of the high-confidence region and a fault classification result set of the low-confidence region , respectively calculate a comprehensive determination result weight of the high-confidence region and a comprehensive classification result weight of the low-confidence region : ， wherein, is the number of high-confidence region signal values, is the number of low-confidence region signal values; S62, according to the comprehensive judgment result weight and the comprehensive classification result weight of the low confidence region The preliminary fault judgment result set of the high confidence region and the fault classification result set of the low confidence region are weighted and fused to generate a comprehensive fault score : ； wherein, is the failure score for the high confidence region, computed directly based on the rule decision result, if is a potential failure, then , otherwise ; is the failure classification probability distribution output for the low confidence region; S63, set the comprehensive fault score threshold comprehensive fault score of each segment matrix q If , the segment is considered to correspond to a fault state, and the final fault judgment result set is generated : ， wherein is a dynamic adjustment coefficient used to comprehensively evaluate the global influence of the segment, S64, record the individual determination results of the high confidence area and the low confidence area, and combine the comprehensive fault determination result set Output the final report, including the running state of each segment, potential fault location, and fault category information.

8. The method of claim 7, wherein the method further comprises: Obtaining a comprehensive fault judgment result set Encode the state information corresponding to each segment index q in the set to generate a fault state encoding matrix E: ， wherein represents the fault state code of the segment q, 1 represents the fault state, and 0 represents the normal state. S82, encode the fault status matrix E with the segmentation matrix index q and the corresponding timestamp correlation, generate a fault trigger signal sequence ; S83, the fault trigger signal sequence input gas relay control module, parse the fault state information therein and the corresponding time to trigger the decision of protection action; when the control module immediately executes the protection action, including power-off operation, alarm triggering or data storage; S84, store the comprehensive fault judgment result set and the analysis result of the fault trigger signal sequence as a historical operation record; S85、According to the running feedback of the gas relay control module, adjust the comprehensive fault score threshold or dynamically adjust the coefficient , optimize the next step of fault judgment process.

9. A failure prediction device for reducing the probability of malfunction of a digital gas relay, comprising a processor and a memory having stored program instructions, characterized in that, The processor is configured to execute the program instructions when running, the fault prediction method for reducing the misoperation probability of the digital gas relay as claimed in any one of claims 1 to 8.

10. A storage medium storing program instructions, characterized in that, The program instructions execute the fault prediction method for reducing the misoperation probability of the digital gas relay as claimed in any one of claims 1 to 8 when running.

Images