Intelligent power failure analysis system and analysis method

Abstract

The invention discloses an intelligent power failure analysis system, which comprises an AI deep learning module, an AI auxiliary preset operation module, a central processing module and an information acquisition module, and is characterized in that the AI deep learning module is connected with the central processing module and the AI auxiliary preset operation module, and the AI auxiliary preset operation module and the information acquisition module are connected with the central processing module; and the AI deep learning module is connected with the big data power failure analysis report library and the constructed and continuously improved transformer substation model. According to the method, deep learning is carried out on a power fault analysis report with big data updated through AI, a transformer substation model which is constructed and continuously improved is combined, fault characteristic quantity is extracted and compared with remote signaling and telemetering data in the normal operation period and the fault period, on one hand, the fault type and the fault position are rapidly analyzed after the fault, and the fault diagnosis accuracy is improved; high-risk faults are removed at the first time, and an analysis report is given; and on the other hand, the system operation data is continuously detected in real time in a non-fault state, so that the aim of preventing electric power accidents in advance is fulfilled.

Description

Technical Field

[0001] The present invention relates to power system fault analysis, and particularly to an intelligent power fault analysis system and an analysis method. Background Art

[0002] After a power fault occurs in a substation power system, the fault should be removed quickly and accurately enough to avoid causing secondary faults or fault expansion. These two fault handling criteria have high requirements for the speed and accuracy of fault analysis. At the same time, comprehensive fault analysis is also of great significance for the subsequent formulation and implementation of abnormal system operation states and fault prevention measures.

[0003] The existing technical solution at present is to collect and send all the teleinformation and telemetry signals collected and sent by the protection devices, sensors, etc. of the substation to the substation automation monitoring system. When a fault occurs, the substation automation monitoring system will output all the fault alarm information, switch position information, switch status information, relay protection information, etc. of the on-site protection devices in the system for the operation personnel to view. The problems with this method are as follows: 1. There are a large number of alarms after a fault occurs. When the communication system configuration is low, it is easy to cause communication congestion and it is impossible to judge the fault type in the first time; 2. Fault analysis depends on the level of operation personnel. Operation personnel with insufficient experience or knowledge cannot conduct timely and effective analysis; 3. The fault alarm information displayed by the substation automation monitoring system depends on the alarm output of the on-site protection device itself. When the setting value of the on-site protection device is set incorrectly, it cannot play the role of alarm and protection; 4. For the critical state that has not reached a fault, a single protection device that has not reached the protection setting value will not start an alarm or a protection action, and it is impossible to achieve a comprehensive analysis of the whole station system. Summary of the Invention

[0004] The technical problem to be solved by the present invention is how to quickly judge the fault type and fault location after a fault occurs in an enterprise substation, so that the operator can perform the next step of processing in time, thereby avoiding the expansion of the fault caused by the extension of the fault state time or misoperation due to wrong judgment.

[0005] To solve the above technical problem, the technical solution adopted by the present invention is: an intelligent power fault analysis system, including: an AI deep learning module, an AI-assisted preset operation module, a central processing module, and an information collection module. The AI deep learning module is connected to the central processing module and the AI-assisted preset operation module, and the AI-assisted preset operation module and the information collection module are connected to the central processing module; the AI deep learning module is connected to the big data power fault analysis report library and the substation model constructed and continuously improved.

[0006] The information acquisition module acquires substation telecontrol signals and substation telemetry signals during normal operation and fault periods.

[0007] The central processing module quickly analyzes the fault type and fault location after a substation fault and executes preset operations. At the same time, it gives the fault analysis result and subsequent treatment suggestions. The central processing module performs real-time detection on the system operation data in the non-fault state of the substation, warns of abnormal operation states that do not reach the fault critical state, and formulates preventive measures.

[0008] The analysis method of an intelligent power fault analysis system of the present invention includes the following steps: S01: The AI deep learning module constructs a substation model and continuously improves it. According to possible fault types, it formulates preset operations to ensure safety and transmits them to the AI-assisted preset operation module; S02: The information acquisition module acquires substation telecontrol signals and telemetry signals during operation and transmits the acquired data to the central processing module; S03-1: In the non-fault state, the central processing module receives the substation telecontrol signals and telemetry signals acquired by the information acquisition module, and interacts and compares them with the AI deep learning module in real time to determine whether there is an abnormal operation state in the system, formulates preventive measures, and proceeds to S04; S03-2: After a fault occurs, the central processing module receives the steady-state changes and transient mutations of the substation telemetry signals acquired by the information acquisition module before and after the fault, and combines the changes in the substation telecontrol signals to initially determine the fault type and fault location; S04: The central processing module transmits the initially judged fault analysis result to the AI deep learning module. The AI deep learning module compares the characteristic quantities of this fault through the substation model and the big data power fault analysis report library, and transmits them to the central processing module; S05: The AI deep learning module and the central processing module interact multiple times to confirm the fault type and location through the coupling degree of the database fault and the characteristic quantities of this fault, and transmit them to the AI-assisted preset operation module. The returned operation instructions are confirmed by the central processing module and then executed; S06 outputs the analysis result of this fault and subsequent treatment suggestions. And updates and optimizes the substation model according to the changes in various transient and steady-state characteristic quantities during this fault.

[0009] An intelligent power fault analysis method and system designed by the present invention adopting the above technical solutions has the following beneficial effects: 1. Based on the substation model, by comparing with the fault characteristic quantities in the fault database, the fault type and location can be accurately judged, which changes the drawback of handing a large amount of data packets to the operators for analysis in the past. In the emergency state of fault handling, the fault type and location are directly output, which is more concise and efficient, while the detailed data packets can be only used for later inductive analysis.

[0010] 2. The fault analysis does not depend on the level of the operators. After a fault occurs, the power fault analysis system can directly cut off the fault according to the preset operations in time. For the subsequent treatment suggestions, they can be directly operated by non-professionals, or the detailed data packets can be analyzed by professionals.

[0011] 3. According to the substation model, this system conducts a secondary protection determination in addition to the alarm action of the on-site protection device, and avoids the occurrence of major safety accidents where the fault cannot be cut off in the case of incorrect settings or refusal to operate of the on-site protection device.

[0012] 4. In the non-fault situation, taking the normal operation situation of the substation as the reference value, detecting the abnormal operation situation, predicting the possible faults, and formulating preventive measures for these faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It shows the block structure schematic diagram of the intelligent power fault analysis system of the present invention; Figure 2 It shows the flow schematic diagram of the intelligent power fault analysis method of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0014] The following specifically describes an intelligent power fault analysis system and analysis method of the present invention with reference to the drawings.

[0015] An intelligent power fault analysis system of the present invention, see Figure 1 , including: an AI deep learning module M03, an AI assisted preset operation module M01, a central processing module M02, and an information collection module M04. The AI deep learning module M03 is connected to the central processing module M02 and the AI assisted preset operation module M01. The AI assisted preset operation module M01 and the information collection module M04 are connected to the central processing module M02. The AI deep learning module M03 is connected to the big data power fault analysis report library and the substation model that is constructed and continuously improved.

[0016] The information acquisition module M04 of the present invention acquires the substation tele-signal quantities and substation tele-measurement quantities during normal operation and fault periods. The central processing module M02 quickly analyzes the fault type and fault location after a substation fault and executes preset operations, and at the same time gives the fault analysis result and subsequent processing suggestions. The central processing module M02 performs real-time detection on the system operation data in the non-fault state of the substation, warns of abnormal operation states that do not reach the fault critical state, and formulates preventive measures.

[0017] An analysis method of an intelligent power fault analysis system of the present invention includes the following steps: S01: The AI deep learning module M03 constructs a substation model and continuously improves it. According to possible fault types, preset operations to ensure safety are formulated and transmitted to the AI-assisted preset operation module M01; (1) Model construction: The AI deep learning module MO3 adopts advanced machine learning algorithms, which can process and analyze a large amount of historical fault data, operation data, and maintenance records to construct an accurate digital twin model of the substation. It includes not only the static parameters of the equipment, such as the rated capacity of the transformer and the opening and closing time of the circuit breaker, but also simulates the dynamic behavior of the equipment, such as the response under different loads and environmental conditions. By integrating an expert system, the model can simulate the decision-making process of experts when facing specific faults, so as to provide decision support when an actual fault occurs; (2) Big data integration: When initializing the substation model, the AI deep learning module M03 integrates big data from different sources, including but not limited to historical telemetry data of the SCADA system, fault records, meteorological environment data, equipment maintenance logs, etc. These data are cleaned, transformed, and loaded into the model through data preprocessing techniques to ensure the quality and consistency of the data, thereby improving the generalization ability and prediction accuracy of the model; (3) Model self-improvement: The model is built with an adaptive learning mechanism that can continuously adjust and optimize model parameters according to the latest operation data and fault cases. This mechanism enables the model to capture subtle changes in substation operation and predict potential fault patterns. Through online learning techniques, it adapts to changes in grid operation conditions. This self-improving mechanism allows the model to automatically adjust when new data arrives, thereby reducing the risk of model obsolescence and maintaining the accuracy and relevance of the model.

[0018] S02: The information acquisition module M04 acquires the tele-signal quantities and tele-measurement quantities during substation operation and transmits the acquired data to the central processing module M02.

[0019] S03-1: If it is in a non-fault state, the central processing module M02 receives the telecontrol signals and telemetry signals collected by the information acquisition module M04, and interacts and compares with the AI deep learning module M03 in real time to determine whether there is an abnormal operating state in the system, formulate preventive measures, and go to S04; (1) Real-time data stream processing: The central processing module MO2 uses stream processing technology to analyze the telecontrol signals and telemetry signals collected by the information acquisition module MO4 in real time. These technologies can process high-speed data streams to ensure immediate analysis when the data arrives, thus achieving real-time monitoring of the substation status; (2) Abnormal detection algorithm: When receiving data from the information acquisition module MO4, the central processing module MO2 uses data synchronization and verification technology to ensure the integrity and accuracy of the data. Any deviation or abnormality in the data will be immediately identified and trigger further analysis. Combining AI algorithms such as autoencoders and isolation forests, the central processing module MO2 can identify abnormal behaviors that do not conform to the model expectations and trigger early warnings. These algorithms can detect abnormal patterns in the data even if these patterns have not appeared in historical data; (3) Generation of preventive measures: Based on big data analysis, the central processing module MO2 can automatically generate preventive measures. These measures may include adjusting load distribution, optimizing operating parameters or performing maintenance in advance to prevent potential failures. At the same time, the central processing module MO2 also includes self-healing control strategies that can automatically adjust system parameters when detecting minor abnormalities to restore the stable operation of the system and reduce the dependence on manual intervention. When detecting potential abnormal operating states that cannot be self-healed, the system will automatically trigger an early warning and provide possible causes and recommended preventive measures. This helps operators take actions in advance to avoid the occurrence of failures.

[0020] S03-2: After a fault occurs, the central processing module M02 receives the steady-state changes and transient mutations of the telemetry measured before and after the fault collected by the information acquisition module M04, and combines the changes in the telecontrol signals of each substation to initially determine the fault type and fault location; (1) Fault feature extraction: Using deep learning technologies such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), the central processing module MO2 extracts fault features from the steady-state changes and transient mutations of the telemetry. These networks can identify complex patterns and time series data, and build a fault feature library that contains the telemetry signal features of various fault types. By matching the real-time telemetry data with the fault feature library, it provides a basis for the preliminary judgment of the fault type and location; (2) Real-time fault diagnosis: Through interaction with the AI deep learning module MO3, the central processing module MO2 can utilize the substation model and real-time telemetry signal changes, adopt advanced fault location algorithms, such as graph theory-based search algorithms, and combine the telemetry signal change information to accurately locate the fault location. This helps to quickly locate the fault location and type and reduce the impact on the power grid. This diagnosis can be based on a rule engine or a machine learning model to achieve fast and accurate fault identification; (3) Fault type classification: Applying classification algorithms, such as support vector machines (SVM) or neural networks, the central processing module MO2 classifies the extracted fault characteristics to determine the specific fault type. At the same time, the fault detection threshold is dynamically set according to historical data and real-time monitoring data to adapt to different operating conditions and load changes, improving the sensitivity and accuracy of fault detection. This classification can help operators understand the nature of the fault and take corresponding measures.

[0021] S04: The central processing module M02 transmits the preliminary fault analysis result to the AI deep learning module M03. The AI deep learning module M03 compares the characteristic quantities of this fault through the substation model and the big data power fault analysis report library and transmits them to the central processing module M02; (1) Deep learning comparison: The AI deep learning module MO3 uses deep learning technology to conduct in-depth comparison and analysis of the fault characteristics transmitted by the central processing module MO2. By training a neural network to identify fault patterns, the AI deep learning module MO3 can provide more accurate fault type and location predictions; (2) Fault case matching: The AI deep learning module MO3 compares the characteristic quantities of this fault with historical fault cases and the substation model. This comparison can be based on similarity metrics to find the historical fault case that best matches the current fault characteristics and provide a reference for fault analysis; (3) AI-assisted decision-making: Through AI algorithms, the AI deep learning module MO3 can assist in decision-making, provide more accurate fault type and location information, and enhance the accuracy of fault analysis. This can include using reinforcement learning to optimize the decision-making process or using Bayesian networks to evaluate the likelihood of different fault hypotheses. The system can recommend the most appropriate handling strategy based on the fault type, location, and severity, thus providing suggestions and guidance for operators to handle faults.

[0022] S05: The AI deep learning module MO3 interacts with the central processing module MO2 multiple times to confirm the fault type and location through the coupling degree of the database fault and the characteristic quantities of this fault, and transmits them to the AI-assisted preset operation module M01. The returned operation instructions are confirmed by the central processing module M02 and then executed; (1) Multi-round dialogue mechanism: A multi-round dialogue mechanism is adopted between the AI deep learning module MO3 and the central processing module MO2 to ensure the accuracy of fault analysis. This mechanism allows the system to collect more information at different stages to refine and confirm the fault analysis results. Through multiple interactions and using big data analysis techniques, the coupling degree between historical fault data and the fault feature quantities of this time is evaluated to confirm the fault type and location. This analysis helps the system understand the relationship between fault features and historical data, thereby improving the accuracy of fault diagnosis; (2) Generation of operation instructions: Based on the confirmed fault information, the AI deep learning module MO3 generates corresponding operation instructions. The operation instructions generated by the AI deep learning module MO3 will be optimized to ensure that the impact on the power grid can be minimized when executed. This includes considering the current state and load demand of the power grid, as well as possible alternative paths. And it is transmitted to the central processing module MO2 for execution through the AI-assisted preset operation module MO1. These instructions can include isolating the fault area, redirecting the power flow, or starting the backup power supply, etc., to minimize the impact of the fault on the power grid. At the same time, after executing the operation instructions, the central processing module MO2 will provide feedback to the AI deep learning module MO3 to evaluate the effect of the instructions. This feedback mechanism helps the system learn and improve the future fault handling process.

[0023] S06 outputs the results of this fault analysis and subsequent processing suggestions. And update and optimize the substation model according to the changes of various transient and steady-state feature quantities during this fault: Fault report generation: The system will output a detailed fault analysis report, including the fault type, location, possible causes, and subsequent processing suggestions. These reports can be automatically generated based on templates to ensure the consistency and integrity of the information; Model optimization and update: The AI deep learning module MO3 updates and optimizes the substation model according to the changes of the transient and steady-state feature quantities of this fault, using machine learning techniques to improve the prediction ability of the model and the accuracy of fault analysis. This optimization can include adjusting model parameters, adding new data features, or retraining the model to adapt to new operating conditions. The knowledge base of the system will be updated according to the latest fault cases and processing results to enrich the knowledge base for fault analysis and processing. This helps improve the intelligence level and efficiency of the system in future fault handling.

[0024] The present invention changes the previous method of analyzing a large amount of data messages provided by a substation automation monitoring system, relying on the experience of operators to analyze the data, and then judging the fault type and location for processing. The AI performs deep learning on the power fault analysis report updated with big data, combines with the substation model constructed and continuously improved, extracts fault feature quantities, and compares them with the remote signaling and telemetry data during normal operation and during the fault. On the one hand, it can quickly analyze the fault type and location after the fault, cut off high-risk faults in the first time according to the preset operations assisted by AI, and give an analysis report for professional or non-professional personnel to analyze and execute; on the other hand, it can continuously detect the system operation data in the non-fault state, give early warnings for abnormal operation states that have not reached the critical state of faults and formulate preventive measures, so as to achieve the purpose of preventing power accidents in advance.

Claims

1. An intelligent power failure analysis system, characterized in that Including: An AI deep learning module, an AI-assisted preset operation module, a central processing module, and an information collection module. The AI deep learning module is connected to the central processing module and the AI-assisted preset operation module, and the AI-assisted preset operation module and the information collection module are connected to the central processing module. The AI deep learning module is connected to a big data power failure analysis report library and a substation model that is constructed and continuously improved.

2. An intelligent power failure analysis system according to claim 1, characterized in that The information collection module collects substation telecontrol signals and substation telemetry signals during normal operation and during a fault.

3. An intelligent power failure analysis system according to claim 1, characterized in that After a substation fault occurs, the central processing module quickly analyzes the fault type and fault location and executes preset operations, and at the same time gives a fault analysis result and subsequent processing suggestions. When the substation is in a non-fault state, the central processing module performs real-time detection on the system operation data, gives an early warning of an abnormal operation state that has not reached the fault critical state, and formulates preventive measures.

4. Analysis method of an intelligent power failure analysis system, characterized in that Including the following steps: S01: The AI deep learning module constructs a substation model and continuously improves it. According to possible fault types, it formulates preset operations to ensure safety and transmits them to the AI-assisted preset operation module. S02: The information collection module collects substation telecontrol signals and telemetry signals during substation operation and transmits the collected data to the central processing module. S03-1: If it is in a non-fault state, the central processing module receives the substation telecontrol signals and telemetry signals collected by the information collection module, and interacts and compares them with the AI deep learning module in real time to determine whether there is an abnormal operation state in the system, formulates preventive measures, and proceeds to S04. S03-2: After a fault occurs, the central processing module receives the steady-state change and transient mutation of the substation telemetry signals collected by the information collection module before and after the fault, and at the same time combines the change of each substation telecontrol signal to initially judge the fault type and fault location. S04: The central processing module transmits the initially judged fault analysis result to the AI deep learning module. The AI deep learning module compares the characteristic quantities of this fault through the substation model and the big data power failure analysis report library and transmits them to the central processing module. S05: The AI deep learning module and the central processing module interact multiple times. The fault type and location are confirmed through the coupling degree of the database fault and the characteristic quantities of this fault, and are transmitted to the AI-assisted preset operation module. The returned operation instructions are confirmed by the central processing module and then executed. S06: Output the analysis result of this fault and subsequent processing suggestions, and update and optimize the substation model according to the changes of various transient and steady-state characteristic quantities during this fault.

Images