Method and system for detecting fault location of intelligent GIS equipment based on cloud-edge connectivity

Abstract

This invention discloses a method for detecting fault locations of intelligent GIS equipment based on cloud-edge collaboration, which relates to the field of computer technology. By utilizing cloud-edge collaboration and combining a digital twin model with anomaly change pattern recognition techniques, this invention achieves comprehensive equipment monitoring, fault prediction, and accurate location. Furthermore, the application of a digital twin model to simulate the interaction relationships between GIS equipment modules is a first in the field of GIS equipment fault location, providing new ideas and methods for solving complex equipment fault problems. Furthermore, a cloud-based large-scale interaction detection model built using a graph neural network algorithm can monitor and analyze GIS equipment interaction data in real time, improving the precision and recall of interaction anomaly detection. Furthermore, by using intelligent algorithms and automatic repair techniques, equipment faults can be automatically repaired after fault location is identified, thereby improving the reliability and stability of the equipment.

Description

[Technical Field]

[0001] The present invention relates to the field of computer technology, and more particularly to a method and system for detecting the location of a fault in an intelligent GIS device based on cloud-edge connectivity. [Background technology]

[0002] Troubleshooting existing GIS equipment typically relies on manual inspection and judgment to locate and handle faults. However, this method has several limitations and drawbacks. First, the complex and diverse structure and function of GIS equipment mean that traditional methods cannot accurately identify hidden fault signals, potentially resulting in faults going undetected and not being addressed in real time. Second, traditional methods typically require significant manpower and time, reducing troubleshooting efficiency and potentially leading to prolonged downtime and repair delays. Finally, existing technologies are insufficient for comprehensive equipment monitoring and fault prediction, and cannot detect potential signs of failure in advance, thus preventing the implementation of preventive maintenance measures. [Overview of the project]

[0003] The purpose of this section is to outline some embodiments of the present invention and to briefly describe some preferred embodiments. This section, the abstract of this application, and the title of the invention may be simplified or omitted to avoid obscuring their purpose, and such simplifications or omissions are not intended to limit the scope of the present invention.

[0004] This invention was proposed in view of the limitations and shortcomings in troubleshooting existing GIS equipment.

[0005] Therefore, the problem that the present invention aims to solve is to overcome the limitations of conventional methods in GIS equipment fault location and processing by applying digital twin models, cloud-edge collaboration ideas, and anomaly change pattern recognition methods, thereby improving the capabilities of equipment monitoring, fault prediction, and accurate location.

[0006] To solve the above technical problems, the present invention provides the following technical solutions.

[0007] According to the first aspect, an embodiment of the present invention provides a method for detecting the location of failures in an intelligent GIS device based on cloud-edge collaboration, which involves constructing a digital twin model in the cloud that includes all modules of the GIS device and their interaction relationships, generating a large amount of failure simulation data by simulating changes in interaction relationships under different failure conditions, extracting interaction data between components, constructing a cloud-based large-scale interaction detection model based on a graph neural network, optimizing the model hyperparameters, and simultaneously deploying the optimized large-scale interaction detection model to a cloud platform, placing sensors locally on the GIS device to collect interaction data of components, constructing an edge-based small-scale interaction localization model to analyze the interaction data, and when an interaction anomaly is detected in the edge-based small-scale interaction localization model, a cloud-based large-scale interaction localization model is deployed. The system includes: requesting a global analysis from an action detection model; a cloud-based large-scale interaction detection model comprehensively determining the fault source location and cause of the failure and feeding the results back to an edge-based small-scale interaction localization model; the edge-based small-scale interaction localization model assisting in local detail determination based on the feedback to achieve accurate fault localization and sending the localization results to the cloud-based large-scale interaction detection model to request correlation analysis; the cloud-based large-scale interaction detection model providing the fault type and root cause based on the localization results; the edge-based small-scale interaction localization model verifying the root cause to complete overall fault localization; and, based on the failure results output from multiple models, querying a repair knowledge base to determine the optimal solution and issuing control commands to automate the repair process, with the system making secondary decisions to ensure fault resolution.

[0008] As a preferred embodiment of the fault location detection method for intelligent GIS equipment based on cloud-edge collaboration described in the present invention, requesting a global analysis from a cloud-based large-scale interaction detection model includes the following steps: an edge-based small-scale interaction localization model continuously detects interaction data between members, and if the interaction data meets the criteria, the edge-based small-scale interaction localization model determines that an interaction anomaly has occurred and takes a snapshot of the interaction data at the moment the anomaly occurred; the edge-based small-scale interaction localization model packages the snapshot of the interaction data and requests a global analysis from a cloud-based large-scale interaction detection model via the network; the cloud-based large-scale interaction detection model receives the data, reconstructs the failure scenario in a digital twin environment, and simulates the propagation and impact of the interaction data to identify the fault source; the cloud-based large-scale interaction detection model comprehensively determines the fault source, conducts experiments and simulations on the fault source, and estimates possible causes of failure; and the cloud-based large-scale interaction detection model feeds back and transmits the estimated fault source and cause of failure to the edge-based small-scale interaction localization model.

[0009] In a preferred embodiment of the fault location detection method for intelligent GIS equipment based on cloud-edge collaboration described in the present invention, it is determined that an interaction anomaly has occurred when the interaction data deviation exceeds μ+3σ and the interaction response time is 2T e The criteria for determining an interaction anomaly are: exceeding a certain threshold, a significant change in the interaction mode, an interruption of the interaction session, or an abnormal change in the statistical characteristics of the interaction data; for interaction data deviation anomalies, the mean μ and standard deviation σ of normal data are calculated to determine the interaction data deviation threshold as μ+3σ, and if the real-time data exceeds the interaction data deviation threshold μ+3σ, it is determined to be an interaction anomaly; and for interaction response time anomalies, the mean response time T eThe interaction response time is calculated and determined to be 2T e This includes determining an interaction anomaly if a certain threshold is exceeded, using time series analysis to determine if a significant change occurs in the interaction mode, analyzing the trends, periodicity, and seasonality of the series to establish a normal interaction mode, comparing real-time interaction data with the normal interaction mode, and determining an interaction anomaly if they do not match, determining if an interaction session is interrupted by setting up a heartbeat detection mechanism for key interaction signals, and determining an interaction anomaly if the heartbeat times out, and determining if an abnormal change occurs in the statistical characteristics of the interaction data by constructing a statistical feature model, detecting whether the real-time data deviates significantly from the established statistical distribution, and determining an interaction anomaly if it deviates significantly.

[0010] As a preferred embodiment of the fault location detection method for intelligent GIS equipment based on cloud-edge collaboration described in the present invention, the edge-based small-scale interaction localization model supports local detailed determination based on feedback to achieve accurate fault localization, which involves the following steps: the edge-based small-scale interaction localization model acquires information on the fault source location fed back from the cloud, monitors this fault location intensively, and simultaneously collects more real-time interaction data; uses correlation analysis to identify the interaction signal with the strongest correlation to the fault, constructs a time-series model of the interaction signal with the strongest correlation to search for subtle anomaly change patterns; traces the signal to identify the signal source, filters it in combination with location information fed back from the cloud, eliminates unlikely components, and accurately identifies the fault location; and localization results and interaction signal data The process includes the steps of: feeding back and transmitting the set to a cloud-based large-scale interaction detection model, requesting the cloud-based large-scale interaction detection model to perform correlation analysis based on global data to determine the failure type and root cause, and building a time-series model of the interaction signal with the strongest correlation to search for subtle anomaly change patterns, namely, obtaining an interaction signal dataset of fault source locations fed back from the cloud; calculating the correlation coefficient between each interaction signal and the failure using Pearson correlation coefficient method for each interaction signal; sorting the correlation coefficients and selecting the top 5 interaction signals from the sorted list as the signals with the strongest correlation to the failure; establishing an ARIMA model based on the interaction signal with the strongest correlation, predicting the correlation signal using the ARIMA model, and simultaneously adjusting the model based on the prediction effect of the ARIMA model; and the model prediction error e t The method includes the steps of using to make a judgment, detecting subtle abnormal change patterns, analyzing the detected abnormal change patterns, and obtaining more information about the occurrence of failures from there.

[0011] As a preferred embodiment of the method for detecting the failure position of an intelligent GIS device based on cloud-edge cooperation described in the present invention, the correlation coefficient ρ XY The specific calculation formula is as follows: JPEG0007845618000001.jpg28150 Here, ρ XY is the correlation coefficient, σ X and σ Y are the standard deviations of variables X and Y, cov represents the covariance of variables X and Y, and the range of possible values of ρ XY is between -1 and 1. The closer the absolute value of ρ XY is to 1, the stronger the linear relationship between the two variables. The closer the absolute value of ρ XY is to 0, the weaker the linear relationship between the two variables. When ρ XY > 0, it indicates that there is a positive correlation between the two variables. That is, when one variable increases, the other variable also increases. When ρ XY <0, it indicates that there is a negative correlation between the two variables. That is, when one variable value increases, the other variable value decreases. When ρ XY = 0, it indicates that the two variables are non-linearly correlated but correlated in other respects. When ρ XY = 1 and ρ XY = -1, it means that the sample points of the two variables X and Y are both on a straight line.

[0012] As a preferred embodiment of the method for detecting the failure position of an intelligent GIS device based on cloud-edge cooperation described in the present invention, the ARIMA model includes a lagged-difference test model, an intercept test model, and a trend test model. The trend test model removes serial correlation by adding a time trend. The specific calculation formula is as follows: JPEG0007845618000002.jpg27150 Here, t is the time variable, X i is an arbitrary time series, εt is white noise, a is a constant term, and βt represents a trend term.

[0013] The actual test starts from the trend test model. If the trend test model is a stationary series, the test stops. Otherwise, the test of the intercept test model continues. If the intercept test model is a stationary series, the test stops. Otherwise, the test continues to check whether the lagged difference test model is stationary. As long as the test result of one of the three models can reject the null hypothesis, the time series is considered stationary. If none of the test results of the three models can reject the null hypothesis, the time series is considered non-stationary.

[0014] As a preferred embodiment of the method for detecting the failure location of an intelligent GIS device based on cloud-edge cooperation described in the present invention, the secondary judgment includes the following steps: that is, querying the repair knowledge base based on the failure result output from the model to find the standard repair solution corresponding to the location determination result; performing a simulation based on the real-time state data and the digital twin model to judge the feasibility of the standard solution, and individually performing optimization and adjustment to form an optimal repair solution; decomposing the optimized repair solution into an ordered control command and issuing the control command to the edge controller; after receiving the command, the edge controller checks the rationality and safety of the command to avoid new failures; the edge controller executes the repair control command to complete the automatic repair process; after the repair is completed, newly performing state detection and diagnosis to confirm whether the failure has been resolved; if the failure has not been resolved, re-executing the location determination analysis until the problem is solved; and recording the current repair event and process and feeding back to the knowledge base for continuous iterative optimization.

[0015] According to a second aspect, an embodiment of the present invention provides a fault location detection system for an intelligent GIS device based on cloud-edge cooperation, constructs a digital twin model including all modules of the GIS device and their interaction relationships in the cloud, and uses a digital twin construction module used to generate a large amount of fault simulation data by simulating changes in interaction relationships in different fault situations, and when an interaction anomaly is detected in an edge-based small-scale interaction location identification model, requests a global analysis from a cloud-based large-scale interaction detection model, and a cloud-based large-scale interaction detection model comprehensively determines the fault source location and the cause of the fault, and an interaction anomaly detection module used to feedback the result to the edge-based small-scale interaction location identification model, and gives a fault type and a root cause based on the location identification result, and verifies the root cause by the edge-based small-scale interaction location identification model to complete overall fault location identification, and an overall fault location identification module used, and after the repair is completed, newly performs state detection and diagnosis, checks whether the fault has been resolved, and if the fault has not been resolved, repeatedly performs location identification analysis until the problem is solved, and a secondary judgment module used.

[0016] According to a third aspect, an embodiment of the present invention provides a computer device including a memory storing a computer program and a processor, and when the computer program commands are executed by the processor, realizes the steps of the fault location detection method for an intelligent GIS device based on cloud-edge cooperation described in the first aspect of the present invention.

[0017] According to a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program commands are executed by a processor, realizes the steps of the fault location detection method for an intelligent GIS device based on cloud-edge cooperation described in the first aspect of the present invention.

[0018] The beneficial effects of this invention are as follows: This invention leverages the idea of ​​cloud-edge collaboration and combines a digital twin model with anomaly change pattern recognition methods to achieve comprehensive equipment monitoring, failure prediction, and accurate localization. Furthermore, applying a digital twin model to simulate the interaction relationships between each module of a GIS device is a first in the field of GIS device failure localization, providing new ideas and methods for solving complex equipment failure problems. In addition, a cloud-based large-scale interaction detection model built using a graph neural network algorithm can monitor and analyze GIS device interaction data in real time, improving the precision and recall of interaction anomaly detection. Moreover, by using intelligent algorithms and automated repair technology, equipment failures can be automatically repaired after the failure location is identified, thereby improving equipment reliability and stability. [Brief explanation of the drawing]

[0019] To more clearly explain the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments are briefly introduced below. Naturally, the drawings in the following description represent only a few embodiments of the present invention, and those skilled in the art can obtain other drawings based on these without requiring any creative effort. In the figures, [Figure 1] This is a flowchart illustrating a method for detecting the location of a fault in intelligent GIS equipment based on cloud-edge connectivity. [Figure 2] This diagram shows the computer equipment used for detecting fault locations in intelligent GIS equipment based on cloud-edge connectivity. [Modes for carrying out the invention]

[0020] To make the above-mentioned objectives, features, and advantages of the present invention clearer and easier to understand, specific embodiments of the present invention will be described in detail below, in conjunction with the accompanying drawings of the specification.

[0021] Many specific details are described below to facilitate a full understanding of the present invention, but the present invention can also be carried out in ways other than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention, and therefore the present invention is not limited by the specific embodiments disclosed below.

[0022] Furthermore, as used herein, “one embodiment” or “example” refers to a specific feature, structure or property that may be included in at least one embodiment of the present invention. The phrase “in one embodiment” appearing elsewhere in this specification does not all refer to the same embodiment, nor are they separate or selectively mutually exclusive embodiments.

[0023] Example 1 Referring to Figures 1 and 2, the first embodiment of the present invention provides a method for detecting the location of a fault in an intelligent GIS device based on cloud-edge collaboration, and includes the following S1 to S7.

[0024] In S1, a digital twin model containing all modules of the GIS equipment and their interaction relationships is built in the cloud, and a large amount of failure simulation data is generated by simulating how the interaction relationships change under different failure conditions.

[0025] Specifically, the process involves collecting design drawings of GIS equipment, including mechanical structure diagrams, circuit principle diagrams, and control logic diagrams; constructing a mechanical model of the equipment, including coordinate systems, motion parameters, and structural constraints, using 3D modeling software based on the structural diagrams; establishing an electrical model using electrical simulation software based on the circuit principle diagrams, and setting parameters for electrical components such as power supplies, sensors, and actuators; constructing a control model using relevant software based on the control logic, creating a control program, setting control cycles and logic decisions, integrating the above submodels to establish a digital twin model, and debugging the normal operating state of the system.

[0026] In S2, interaction data between components is extracted, a cloud-based large-scale interaction detection model based on a graph neural network is constructed, the model's hyperparameters are optimized, and the optimized large-scale interaction detection model is simultaneously deployed to the cloud platform.

[0027] Specifically, the process involves extracting interaction features between components, including timing data such as voltage, current, temperature, and vibration, from normal and faulty samples collected by the digital twin; constructing a connection topology diagram between components; labeling the direction of information and energy flow; building a model using a graph neural network algorithm, with components as nodes and interactions as edges; adjusting hyperparameters such as the number of network layers and node types to design a network structure capable of handling temporal relationships; constructing a graph format of the sample set; configuring interaction features as node vector inputs; performing model training to minimize the classification loss function; obtaining an interaction anomaly detection model; evaluating metrics such as the model's classification precision and recall on a test set; iteratively adjusting parameters; and comparing different structures to obtain the optimal hyperparameter configuration. An optimized large-scale interaction detection model is deployed on a cloud server and packaged as a prediction API interface. The cloud-based large-scale interaction detection model receives interaction data uploaded in real time from edge-based small-scale interaction localization models, enabling the prediction of interaction anomalies.

[0028] In S3, sensors are placed locally on the GIS equipment to collect interaction data between components, and an edge-based small-scale interaction localization model is built to analyze the interaction data.

[0029] Preferably, the placement points of the main components are determined according to the structural diagram of the GIS equipment, appropriate sensors such as current, voltage, and vibration sensors are selected, a sensor distribution plan is formulated, various sensors are placed in their corresponding locations, the range and accuracy parameters of the sensors are set, and a driver program is created to collect real-time monitoring data from each sensor.

[0030] Furthermore, an edge-based small-scale interaction localization model is constructed at the edges of the equipment to enable aggregation and analysis of collected data, a topology connection structure is constructed between components, a link model of information and control flow is set up, an interaction monitoring algorithm is designed, real-time interaction data between components is analyzed, and if the interaction data exceeds the judgment criteria, it is judged as an anomaly, a localization analysis program is activated, the topology structure and real-time data are integrated to perform fault isolation and localization estimation, the faulty component is accurately identified, and the results are output.

[0031] In S4, when an interaction anomaly is detected by the edge-based small-scale interaction localization model, it requests a global analysis from the cloud-based large-scale interaction detection model. The cloud-based large-scale interaction detection model comprehensively determines the fault source location and cause of the failure, and feeds the results back to the edge-based small-scale interaction localization model.

[0032] Specifically, this includes the following steps S4.1 to S4.5.

[0033] In S4.1, the edge-based small-scale interaction localization model continuously detects interaction data between members. If the interaction data meets the criteria, the edge-based small-scale interaction localization model determines that an interaction anomaly has occurred and takes a snapshot of the interaction data at the moment the anomaly occurred.

[0034] Specifically, the interaction data deviation exceeds μ+3σ, and the interaction response time is 2T. eThe determination is made based on the following criteria: exceeding a certain threshold, a significant change in the interaction mode, interruption of the interaction session, or an abnormal change in the statistical characteristics of the interaction data.

[0035] Specifically, for interaction data deviation anomalies, the mean μ and standard deviation σ of normal data are calculated to determine the interaction data deviation threshold as μ+3σ. If the real-time data exceeds the interaction data deviation threshold μ+3σ, it is determined to be an interaction anomaly. For interaction response time anomalies, the mean response time T is determined. e The interaction response time is calculated and determined to be 2T e If it exceeds this value, it is determined to be an interaction anomaly, and at this point, the mean response time T e The average value of the historical response time is used to determine if a significant change occurs in the interaction mode. A normal interaction mode is established by analyzing the characteristics of the series, such as trends, periodicity, and seasonality. Real-time interaction data is then compared with the normal interaction mode. If they do not match, it is determined to be an interaction anomaly. Interaction session interruptions are determined by setting up a heartbeat detection mechanism for key interaction signals. If the heartbeat is a time attack, it is determined to be an interaction anomaly. Abnormal changes in the statistical characteristics of the interaction data are determined by constructing a statistical feature model. It is detected whether the real-time data deviates significantly from the established statistical distribution. If it deviates significantly, it is determined to be an interaction anomaly.

[0036] In S4.2, a snapshot of the interaction data is packaged, and a global analysis is requested via the network to a cloud-based, large-scale interaction detection model.

[0037] In S4.3, a cloud-based large-scale interaction detection model receives data, reconstructs this failure scenario in a digital twin environment, and simulates the propagation and impact of interaction data to identify the source of the failure.

[0038] In S4.4, the cloud-based large-scale interaction detection model comprehensively determines the fault source location and estimates possible causes of failure by conducting experiments and simulations on that location.

[0039] Preferably, after identifying the fault source, experiments and simulations are performed on the fault source to identify the problem and objective to be solved, clarify the fault source and interaction process to be experimented and simulated, design an appropriate experimental scheme according to the characteristics of the fault source and interaction process, simulate specific impact environments, input conditions and other parameter settings according to the requirements of the fault source and interaction process, record and observe experimental data and phenomena and changes, collect data generated during the experiment using equipment such as sensors and recorders, analyze and process the collected experimental data to find patterns, trends and correlations, find what is suspected to be the cause of the failure or abnormal operation by observing and analyzing the data, and verify and confirm the suspected cause of the failure based on the experimental results.

[0040] In S4.5, the estimated fault location and cause are fed back and transmitted to an edge-based small-scale interaction localization model.

[0041] In S5, the edge-based small-scale interaction localization model assists in local detail determination based on feedback to achieve accurate fault localization, and sends the localization results to the cloud-based large-scale interaction detection model to request correlation analysis.

[0042] Specifically, this includes the following steps S5.1 to S5.4.

[0043] In S5.1, the edge-based small-scale interaction localization model acquires information about fault source locations fed back from the cloud, focuses its monitoring on these fault locations, and simultaneously collects more real-time interaction data.

[0044] In S5.2, correlation analysis is used to identify the interaction signal with the strongest correlation to the failure, and a time-series model of the interaction signal with the strongest correlation is constructed to search for subtle anomaly change patterns.

[0045] Specifically, this includes the following steps S5.2.1 to S5.2.5.

[0046] In S5.2.1, a dataset of interaction signals from the fault source is obtained from the cloud.

[0047] In S5.2.2, the correlation coefficient between each interaction signal and the fault is calculated using the Pearson correlation coefficient method.

[0048] Specifically, the correlation coefficient ρ XY The specific calculation formula is as follows: JPEG0007845618000003.jpg28150 Here, ρ XY σ is the correlation coefficient. X and σ Y represents the standard deviation of variables X and Y, and cov represents the covariance of variables X and Y.

[0049] Note, ρ XY The range of possible values ​​for ρ is between -1 and 1. XY The closer the absolute value of is to 1, the stronger the linear relationship between the two variables becomes, and ρ XY The closer the absolute value is to 0, the weaker the linear relationship between the two variables becomes, ρ XY When >0, it indicates a positive correlation between the two variables, that is, as one variable increases, the other variable also increases, ρ XY When <0, it indicates a negative correlation between the two variables; that is, as the value of one variable increases, the value of the other variable decreases, and ρ XY When = 0, it indicates that the two variables are nonlinearly correlated, but are correlated in other respects, and ρ XY =1 and ρ XY When = -1, it means that the two X and Y can be nicely described by a linear equation, and all the sample points lie on a single straight line.

[0050] In S5.2.3, the correlation coefficients are sorted, and the top five interaction signals from the sorted list are selected as the signals with the strongest correlation to the fault.

[0051] Specifically, all correlation coefficients calculated in S5.2.2 are obtained, and each interaction signal and its corresponding correlation coefficient are recorded in a list. The interaction signals in the list are sorted in descending order of their correlation coefficients, showing that a higher correlation coefficient indicates a stronger correlation between the signal and the fault. The top five interaction signals from the sorted list are selected as the signals with the strongest correlation to the fault, and further analysis and research are performed on the selected interaction signals to determine their relationship with the fault.

[0052] In S5.2.4, an ARIMA model is established based on the interaction signal with the strongest correlation, the correlation signal is predicted using the ARIMA model, and the model is simultaneously adjusted based on the predictive effect of the ARIMA model.

[0053] Specifically, time series data of the signal with the strongest correlation are collected, the stationarity of the time series is tested using ADF, if the time series is not stationary, a stationary series is obtained by difference processing, and a white noise test is performed on the stationary series, if the stationary series is white noise, the p and q orders of the ARIMA model are determined, the values ​​of p and q are initially determined using ACK (autocorrelation) and PAC (partial correlation) plots, the ARIMA(p, d, q) model is constructed, the model parameters are estimated using the maximum likelihood method, the optimal model is determined using AIC and BIC criteria, the p and q parameters are adjusted, and the model predicted values ​​are obtained. Generate JPEG0007845618000004.jpg2316, and predict the error e t Calculate the prediction error e t If the results do not match the white noise, adjust the modeling parameters and reconstruct the ARIMA model.

[0054] The time series stationarity test using ADF is implemented using three models. Here, the lag period is added to the lagd difference test model to remove serial correlation, and the specific calculation formula is as follows: JPEG0007845618000005.jpg36150 Here, t is the time variable, X i represents an arbitrary time series, and εt represents white noise.

[0055] By adding an intercept term to the intercept test model and removing serial correlation, the specific calculation formula is as follows: JPEG0007845618000006.jpg31150 Here, t is the time variable, X i represents an arbitrary time series, εt is white noise, and a represents a constant term.

[0056] By adding a time trend to the trend testing model and removing serial correlation, the specific calculation formula is as follows: JPEG0007845618000007.jpg27150 Here, t is the time variable, X i εt represents an arbitrary time series, εt is white noise, a is the constant term, and βt is the trend term.

[0057] Furthermore, the actual testing begins with the trend test model. If the trend test model is a stationary series, the test is stopped; otherwise, the test of the intercept test model continues. If the intercept test model is a stationary series, the test is stopped; otherwise, the test of whether the raged difference test model is stationary continues. There are also corresponding critical value tables for each model test. The time series is considered stationary if the test result of one of the three models rejects the null hypothesis. If none of the test results of the three models reject the null hypothesis, the time series is considered non-stationary.

[0058] The original assumption was that H0:δ=0, meaning that there is one unit root.

[0059] In S5.2.5, the model's prediction error e tIt uses this method to make judgments, detect subtle abnormal change patterns, analyze the detected abnormal change patterns, and obtain more information about the occurrence of failures from them.

[0060] Specifically, the prediction error e between the model's predicted value and the actual observed value. t Calculate the prediction error e t A white noise test was performed on the predicted error e t If it matches the white noise, it indicates that the model fits well; otherwise, it indicates that there is an abnormal change and the prediction error e t If it is detected that the signal does not match the white noise, an anomaly pattern recognition is performed. The residual time series diagram is observed to determine the location, time, and duration of the anomaly, and at the same time, the degree of deviation of the residual is calculated to determine the degree of the anomaly. The recognized anomaly pattern is then analyzed to determine whether or not it is a system failure signal.

[0061] In S5.3, the signal is traced to identify the signal source, and this information is combined with part information fed back from the cloud for filtering, eliminating unlikely components and accurately pinpointing the fault location.

[0062] Specifically, based on the detected fault signal and sensor placement, possible sources of the abnormal signal are preliminaryly determined. The possible sources of the abnormal signal are further narrowed down by referring to the operating characteristics of different components according to the amplitude size and frequency characteristics of the abnormal signal fluctuations. Temporary sensors are installed in specific locations to collect more correlated signal data, and the behavior of the abnormal signal at these sensors is observed to pinpoint the exact source of the abnormal signal. The identified fault signal source is compared with the historical failure site records in the cloud platform database to filter for matching failure sites. This is combined with the failure site probability analysis results fed back from the expert experience system to exclude failure sites deemed unlikely by the system. The consistency of the integrated multiple model results is compared to determine the final failure site. The final failure site is inspected on-site to confirm the actual failure situation and verify the accuracy of the failure location. The relationship between the failure site and the fault signal source is labeled and fed back to the cloud platform to continuously improve the database and enhance subsequent location accuracy.

[0063] In S5.4, the localization results and interaction signal dataset are fed back and transmitted to a cloud-based large-scale interaction detection model, which is then requested to perform a correlation analysis based on global data to determine the failure type and root cause.

[0064] In S6, a cloud-based large-scale interaction localization model provides fault types and root causes based on localization results, and an edge-based small-scale interaction localization model verifies the root causes to complete overall fault localization.

[0065] Specifically, this includes the following steps S6.1 to S6.4.

[0066] In S6.1, the cloud-based large-scale interaction detection model receives accurate fault location data fed back from the edge-based small-scale interaction localization model and constructs the associated topology subgraph of this location on the digital twin platform.

[0067] S6.2: By simulating propagation effects under different failure conditions, the most likely failure pattern is comprehensively analyzed and determined, and the appropriate failure type is determined based on the failure pattern.

[0068] Specifically, the system simulates different types of failures in a subgraph, observes the failure propagation effects, analyzes the propagation effects of different failure patterns under given conditions for the failure location, determines the failure pattern most likely to match the actual situation, and then determines the appropriate failure type based on the failure pattern database.

[0069] S6.3: Use complex system analysis techniques to determine the most likely failure chain, analyze the failure chain to obtain the root cause of the failure, and feed back the estimated failure type and root cause to an edge-based small-scale interaction localization model.

[0070] Preferably, the critical locations of the failure site in the topology diagram are analyzed to construct possible failure chains, the most likely failure chain is simulated and determined based on failure propagation, and degradation analysis is performed along the failure chain to estimate the most likely root cause of the failure.

[0071] S6.4: After receiving a cause, the edge-based small-scale interaction localization model checks whether real-time monitoring data supports that cause.

[0072] Specifically, if it is found that the real-time data does not match the cause, the problem to be analyzed will be provided again, the cloud-edge collaborative analysis will be rerun, and if the data validation is successful, the given fault location results will be confirmed to complete the overall fault location.

[0073] In S7, based on the failure results output from multiple models, the system queries the repair knowledge base to determine the optimal solution and issues control commands to automate the repair process. The system then makes a secondary decision to ensure the failure is resolved.

[0074] Furthermore, based on the failure results output from the model, the system queries a repair knowledge base to find a standard repair solution corresponding to the location result. The knowledge base compiles repair strategies for various types of failures based on historical cases. Simulations are performed based on real-time status data and a digital twin model to determine the feasibility of the standard solution, and the optimal repair solution is formed by individually optimizing and adjusting it. The optimized repair solution is broken down into sequential control commands and issued to the edge controller. These control commands include switching, restarting, and reconfiguring equipment modules. After receiving the commands, the edge controller checks the rationality and safety of the commands to avoid new failures, and then executes the repair control commands to complete the automated repair process.

[0075] Furthermore, after the repair is completed, a new state detection and diagnosis are performed to confirm whether the fault has been resolved. If the fault has not been resolved, the localization analysis is performed again until the problem is resolved, the repair event and process are recorded, and the results are fed back into the knowledge base for continuous iterative optimization.

[0076] Furthermore, this embodiment further provides an intelligent GIS equipment failure location detection system based on cloud-edge collaboration, including: a digital twin construction module used to generate a large amount of failure simulation data by constructing a digital twin model in the cloud that includes all modules of the GIS equipment and their interaction relationships, and simulating changes in interaction relationships under different failure conditions; an interaction anomaly detection module used to request a global analysis from a cloud-based large-scale interaction detection model when an interaction anomaly is detected in the edge-based small-scale interaction localization model, which then comprehensively determines the fault source location and cause of the failure and feeds the results back to the edge-based small-scale interaction localization model; a global failure localization module used to determine the failure type and root cause based on the localization results, and to verify the root cause by the edge-based small-scale interaction localization model to complete the overall failure localization; and a secondary determination module used to perform a new state detection and diagnosis after repair is completed to confirm whether the failure has been resolved, and if not, to repeat the localization analysis until the problem is resolved.

[0077] This embodiment further provides a computer system applicable to a fault location detection method for intelligent GIS equipment, the computer system comprising memory and a processor, the memory being used to store computer-executable commands and the processor being used to execute computer-executable commands, thereby realizing the fault location detection method for intelligent GIS equipment provided in the above embodiment.

[0078] The computer equipment is not a terminal but includes a processor, memory, communication interface, display, and input devices connected via a system bus. Here, the processor of the computer equipment is used to provide calculation and control functions. The memory of the computer equipment includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides the operating environment for the operating system and computer programs in the non-volatile storage medium. The communication interface of the computer equipment is used to communicate with an external terminal by wired or wireless means, and the wireless method can be implemented by Wi-Fi, operator network, NFC (Near Field Communication), or other technology. The display of the computer equipment may be a liquid crystal display or an electronic ink display, and the input devices of the computer equipment may be a touch layer covered by the display, buttons provided on the housing of the computer equipment, a trackball or touchpad, or an external keyboard, touchpad or mouse.

[0079] This embodiment further provides a storage medium on which a computer program is stored, and when the program is executed by a processor, the fault location detection method for the intelligent GIS device presented in the above embodiment is realized.

[0080] In summary, this invention leverages the idea of ​​cloud-edge collaboration and combines a digital twin model with anomaly change pattern recognition techniques to achieve comprehensive equipment monitoring, failure prediction, and accurate localization. Furthermore, applying a digital twin model to simulate the interaction relationships between each module of a GIS device is a first in the field of GIS device failure localization, providing new ideas and methods for solving complex equipment failure problems. Moreover, a cloud-based large-scale interaction detection model built using graph neural network algorithms can monitor and analyze GIS device interaction data in real time, improving the precision and recall of interaction anomaly detection. In addition, by using intelligent algorithms and automated repair techniques, equipment failures can be automatically repaired after fault localization, thereby improving equipment reliability and stability.

[0081] Example 2 Referring to Figures 1 and 2, a second embodiment of the present invention provides a fault location detection method for intelligent GIS equipment based on cloud-edge collaboration, and scientific verification was conducted through economic effect calculations and simulation experiments to verify the beneficial effects of the present invention.

[0082] Specifically, we collected GIS design drawings, built a simulation model of them using a digital twin, and simulated changes in interaction relationships under different failure conditions to generate a large amount of failure simulation data. From this data, we collected 100 sets of failure data and normal data, some of which are shown in Tables 1 and 2.

[0083] Table 1: Normal Data Table for Some GIS Equipment JPEG0007845618000008.jpg41150 Table 2 Failure Data Table for Some GIS Equipment Furthermore, using the data from Tables 1 and 2, sensor output and actuator input are extracted as features, and a cloud-based large-scale interaction detection model based on a graph neural network is constructed. Sensors are installed at the edges of the equipment, normal operation data is collected in real time, and an edge-based small-scale interaction localization model is constructed, and if the sensor output exceeds a threshold, it is judged to be abnormal.

[0084] Furthermore, the edge-based small-scale interaction localization model detects that sensor A has an abnormal output of 11.8V and sends the data to the cloud platform. The cloud-based large-scale interaction detection model analyzes this data using a digital twin and determines that sensor A is the source of the failure. The edge-based small-scale interaction localization model collects 100 sets of sensor A data, detects fluctuations of 11-13V in the output, tests the stationarity of the time series using ADF, and combines this with the ADF distribution critical value table to determine that the time series is stable and that the stable series is white noise. Based on this, the ARIMA model is constructed, detecting that the error does not fit the white noise, and then anomaly change pattern recognition is performed to confirm that sensor A is faulty. The ADF distribution critical value table is shown in Table 3.

[0085] Table 3 ADF distribution critical value table Furthermore, the cloud-based large-scale interaction detection model receives the localization results, simulates fault propagation in the digital twin, determines that the fault type is deviation drift of sensor A, queries the repair knowledge base according to the deviation drift, finds a standard repair solution corresponding to the deviation drift, simultaneously performs simulations based on real-time state data and the digital twin model to verify the feasibility of the standard solution, individually optimizes and adjusts to form the optimal repair solution, breaks down the optimal repair solution into sequential control commands, issues them to adjust the calibration parameters of sensor A, performs a secondary check after the repair is completed to confirm that the fault has been resolved.

[0086] Furthermore, Table 4 shows a comparison between the present invention and conventional GIS equipment for detection and location determination.

[0087] Table 4 Comparison of the present invention with conventional GIS equipment detection and location determination. Preferably, as can be seen from the results in Table 4, the present invention has significant advantages in many aspects, including fault detection accuracy, localization accuracy, response time, repair time, and maintenance costs. The cloud-edge collaboration-based detection and localization idea adopted in the present invention improves analysis and decision-making capabilities through digital twins and multi-source data fusion, resulting in fault detection accuracy and localization accuracy being improved by 11% and 10%, respectively, compared to conventional methods.

[0088] The response time and repair time of the present invention are more advantageous than those of conventional methods, resulting in a significant reduction in downtime for faulty equipment, a substantial improvement in system availability, and a 23% reduction in maintenance costs compared to conventional methods, which do not yield significant benefits.

[0089] The above-described embodiments are merely for illustrating the technical solutions of the present invention and do not limit them. While the present invention has been described in detail with reference to preferred embodiments, those skilled in the art can modify or substitute equivalents to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and such modifications or substitutions should be included within the claims of the present invention.

Claims

1. By building a digital twin model in the cloud that includes all modules of GIS equipment and their interaction relationships, and by simulating changes in interaction relationships under different failure conditions, a large amount of failure simulation data can be generated. This involves extracting interaction data between components, constructing a cloud-based large-scale interaction detection model based on a graph neural network, optimizing the model's hyperparameters, and simultaneously deploying the optimized large-scale interaction detection model to a cloud platform. This involves placing sensors locally on GIS equipment to collect interaction data of components, and then constructing an edge-based small-scale interaction localization model to analyze the interaction data. When an interaction anomaly is detected in the edge-based small-scale interaction localization model, it requests a global analysis from the cloud-based large-scale interaction detection model. The cloud-based large-scale interaction detection model comprehensively determines the fault source location and cause of the failure, and feeds the results back to the edge-based small-scale interaction localization model. The edge-based small-scale interaction localization model supports local detail determination based on feedback to achieve accurate fault localization, and transmits the localization results to a cloud-based large-scale interaction detection model to request correlation analysis. A cloud-based large-scale interaction localization model provides fault types and root causes based on localization results, and an edge-based small-scale interaction localization model verifies the root causes to complete overall fault localization. A method for detecting the location of a fault in an intelligent GIS device based on cloud-edge collaboration, characterized in that the system queries a repair knowledge base to determine the optimal solution based on fault results output from multiple models, issues control commands to automate the repair process, and makes a secondary decision to ensure the fault is resolved.

2. The secondary judgment described above involves the following steps, namely: The steps include querying a repair knowledge base based on the failure results output from the model to find a standard repair solution corresponding to the location result, The process involves performing simulations based on real-time state data and a digital twin model to determine the feasibility of standard solutions, and then individually optimizing and adjusting them to form the optimal repair solution. The steps include: breaking down the optimized repair solution into sequential control commands and issuing the control commands to the edge controller; After receiving a command, the edge controller performs a step to check the reasonableness and safety of the command in order to avoid further failures. The edge controller executes repair control commands and completes the automated repair process. After the repair is complete, a new condition detection and diagnosis is performed to confirm whether the malfunction has been resolved. If the malfunction is not resolved, the localization analysis will be run again until the problem is resolved. A method for detecting the location of a fault in an intelligent GIS device based on cloud-edge collaboration, according to claim 1, characterized by comprising the step of recording the current repair event and process and feeding it back into a knowledge base for continuous iterative optimization.

3. A computer system comprising a computer program stored in a processor, wherein when the processor executes the computer program, it realizes the steps of the method for detecting the location of an intelligent GIS device fault based on cloud-edge collaboration as described in claim 1.

4. A computer-readable storage medium on which a computer program is stored, characterized in that when the computer program is executed by a processor, it realizes the steps of the method for detecting the location of an intelligent GIS device based on cloud-edge collaboration as described in claim 1.

Images