Real-time fault detection and automatic repair system for power grid

Abstract

The present application relates to a kind of power grid real-time fault detection and automatic repair system, including multilevel perception module, data processing module, automation repair module and monitoring and log module.Multilevel perception module is through the deployment of multiple types of sensors, collects voltage, current, frequency and temperature and other data in the operation of power grid, and it is sent to data processing module.Data processing module is analyzed to the acquisition data by embedded intelligent algorithm, identifies potential fault and generates fault diagnosis result.Automatic repair module executes the repair operation of isolating fault area, reconfiguring power grid topology structure and recovering normal power supply according to fault diagnosis result.Monitoring and log module records power grid operating state, fault diagnosis and repair process, and provides fault analysis and performance evaluation function through user interface.The system realizes the real-time detection of power grid fault, accurate positioning, automatic repair and comprehensive monitoring of operating data, significantly improves the safety and reliability of power grid operation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric power, in particular to a real-time fault detection and automatic repair system for power grid. BACKGROUND

[0002] In the prior art, various faults may occur in the operation of the power grid, including short circuit, overload and device anomaly, etc. In order to improve the reliability of the power grid, existing systems collect power grid operation data through sensors and analyze the state thereof, and some systems can locate faults and handle them through manual or semi-automatic methods. In addition, some intelligent systems can provide basic monitoring functions for power grid operation, but are often limited to data recording after the occurrence of faults and the handling of single faults.

[0003] However, the prior art still has obvious problems, mainly manifested in the insufficient real-time performance of fault detection, the insufficient accuracy of diagnosis results and the low degree of automation of the repair process. The existing systems have limitations in identifying fault types and evaluating the influence range, and the topology reconstruction and repair operation of the power grid often require manual intervention, which is difficult to meet the demand of modern power grid for rapid recovery and high reliability.

[0004] In order to solve the above problems, the present application provides a new real-time fault detection and automatic repair system for power grid. SUMMARY

[0005] The present application provides a real-time fault detection and automatic repair system for power grid to improve the safety and reliability of power grid operation.

[0006] The present application provides a real-time fault detection and automatic repair system for power grid, comprising:

[0007] A multi-level perception module is configured to collect operation data of the power grid through multiple types of sensors and send the collected operation data to a data processing module, wherein the operation data includes voltage, current, frequency and temperature.

[0008] A data processing module is configured to receive the operation data provided by the multi-level perception module, analyze the operation data through embedded intelligent algorithms to identify potential faults, and generate a fault diagnosis result when a fault is detected, wherein the fault diagnosis result includes fault type, fault location and influence range.

[0009] An automatic repair module is configured to perform repair operations according to the fault diagnosis result generated by the data processing module, wherein the repair operations include isolating the fault area, reconstructing the topology structure of the power grid and restoring normal power supply.

[0010] A monitoring and logging module is configured to record the power grid operation state, fault diagnosis result and repair operation process, generate a historical data report, and provide fault analysis and operation performance evaluation through a user interface.

[0011] Further, the multi-level perception module is specifically configured to:

[0012] Voltage sensors and current sensors are deployed at substations, transmission and distribution lines and load ends to collect voltage and current data of each node in real time, wherein the voltage sensors are configured to detect the instantaneous value of node voltage, and the current sensors are configured to monitor the amplitude and direction of current;

[0013] Temperature sensors are installed at high-load devices and key line positions to detect the temperature rise of device housings and lines at regular intervals, and the temperature data is transmitted to the data processing unit through a wireless communication module;

[0014] A frequency acquisition device is integrated into the main substation to continuously sample the change of system frequency, capture the frequency deviation caused by load fluctuation or fault, and transmit all collected data to the data processing module after encapsulation.

[0015] Further, the automatic repair module is specifically configured to:

[0016] Based on the fault diagnosis result, a circuit breaker control unit is started to isolate the identified fault area, wherein the isolation includes physically disconnecting the fault line or device from the normal power supply part;

[0017] The power grid topology of the non-fault area is calculated, a new power supply path is generated by enabling backup power supply, backup line or redistributing existing load, and instructions are sent to related switching devices through the communication module to execute topology reconstruction;

[0018] After confirming the completion of topology adjustment, the voltage level and power supply power are adjusted through the transformer to gradually restore the power supply of the normal load area.

[0019] Further, the monitoring and logging module is specifically configured to:

[0020] Real-time operation data, including voltage, current, frequency and temperature parameters, are obtained through a state acquisition unit, abnormal data are recorded when a fault occurs, and operation state snapshots before and after the fault are generated;

[0021] During the fault repair process, the specific steps and execution time of isolation operation, topology adjustment and power supply restoration are recorded in real time;

[0022] After the repair is completed, a complete log report including power grid operation state data, fault cause analysis and repair process is generated, which is displayed to the operator through a visual interface and stored.

[0023] Further, the intelligent algorithm used by the data processing module is implemented using the following steps:

[0024] The operation data collected by the multi-level perception module is decomposed using wavelet transform, and low-frequency components are extracted by removing high-frequency noise; sliding window time series analysis is performed on the low-frequency components to extract characteristic data reflecting the operation state of the power grid, and power grid characteristic data is generated; wherein the extracted characteristic data includes current waveform change rate, voltage unbalance degree, and frequency offset characteristics;

[0025] The characteristic data is input into a pre-trained support vector machine classifier, the classification boundary is dynamically optimized based on historical samples and real-time load data, and the diagnosis result of the fault type is generated; according to the diagnosis result, the abnormal operation parameters related to the fault type are marked, and the fault type includes short circuit, overload, line breakage or equipment abnormality;

[0026] According to the diagnosis result of the fault type and the marked abnormal operation parameters, the physical structure of the power grid is backtracked layer by layer, and the source of voltage or current change is analyzed based on the abnormal current propagation path; according to the analysis result of the source of voltage or current change, and combined with the operation characteristics of the equipment and the line, the specific location of the fault is determined, and the fault positioning result including the fault equipment or line identification is generated;

[0027] Based on the fault positioning result, combined with the connection relationship of the equipment in the power grid and the real-time load state, the influence range of the fault is analyzed by level-by-level propagation, and the operation state change of the substation, transmission line and load equipment and the affected area are gradually calculated; when the influence is weakened to a pre-set negligible range, the detailed information of the affected area is generated, including equipment identification, operation parameter change value and specific influence range.

[0028] Further, the use of wavelet transform to decompose the operation data collected by the multi-level perception module, and extract low-frequency components by removing high-frequency noise, includes:

[0029] According to the following formula (1) and formula (2), the wavelet base functions ψ j,k (t) and φ j,k (t) are defined as:

[0030]

[0031] Wherein, t represents the sampling time; j represents the scale parameter; k represents the time shift parameter;

[0032] ψ(t) and φ(t) are the base functions of the wavelet function, as shown in the following formula (3) and formula (4):

[0033] ψ(t) = exp(-αt 2)cos(2πβt) (3)

[0034] φ(t)=exp(-αt 2 ) (4)

[0035] wherein, alpha is a frequency offset adjustment parameter; beta is a waveform adjustment parameter; t represents a sampling time;

[0036] According to the following formula (5) and (6), the operation data x(t) is multi-scale decomposed to generate high frequency component D j (t) and low frequency component A j (t):

[0037]

[0038] According to the following formula (7), the high frequency component threshold T is calculated:

[0039]

[0040] wherein, eta is a proportional factor set based on statistical characteristics of power grid operation data; N is the number of sampling points;

[0041] The high frequency component of |D j (t)|>T is set to zero to remove noise to obtain the denoised high frequency component;

[0042] By inverse wavelet transform on the denoised high frequency component and low frequency component, the low frequency component reflecting the power grid operation state is reconstructed and extracted as the key feature input to describe the power grid state.

[0043] The present application has the following beneficial technical effects:

[0044] (1) The present application can accurately identify potential faults in power grid operation by embedding intelligent algorithms in the data processing module, combined with multi-dimensional operation data collected by the multi-level perception module, to generate diagnostic results including fault type, fault location and impact range, providing reliable basis for rapid repair measures.(2) The automatic repair module can independently execute isolation operation of the fault area according to the diagnostic results, quickly reconstruct the power grid topology and restore normal power supply, significantly reducing the power outage time and the need for manual intervention, improving the operation reliability and maintenance efficiency of the power grid.(3) Through the monitoring and log module, the system can record the operation state of the power grid, fault diagnosis results and repair operation process in detail, generate historical data reports, and provide intuitive fault analysis and performance evaluation through the user interface, supporting the operation optimization and maintenance decision of the power grid.(4) The system realizes real-time response of the whole process from fault detection to repair, and has the ability to quickly handle sudden faults. At the same time, the modular design makes each functional unit independent and mutually cooperative, facilitating flexible configuration and expansion according to the scale of the power grid and application requirements. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 is a schematic diagram of an electric power grid real-time fault detection and automatic repair system provided by the first embodiment of the present application. DETAILED DESCRIPTION

[0046] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application. However, the present application can be practiced in a variety of ways beyond those described herein, and it is understood that it is intended to cover all such variations as fall within the scope of the present application. Therefore, the present application is not limited to the specific embodiments disclosed below, but only by the claims.

[0047] The first embodiment of the present application provides an electric power grid real-time fault detection and automatic repair system. Please refer to Figure 1 , which is a schematic diagram of the first embodiment of the present application. The following will be described in detail Figure 1 The first embodiment of the present application provides an electric power grid real-time fault detection and automatic repair system. Please refer to Figure 1 , which is a schematic diagram of the first embodiment of the present application. The following will be described in detail

[0048] The electric power grid real-time fault detection and automatic repair system comprises a multi-level perception module 101, a data processing module 102, an automatic repair module 103 and a monitoring and log module.

[0049] The multi-level perception module 101 is used to collect the operation data of the electric power grid through multiple types of sensors, and send the collected operation data to the data processing module; wherein the operation data includes voltage, current, frequency and temperature.

[0050] The multi-level perception module 101 aims to efficiently and timely collect key parameters of the operation of the electric power grid through multiple types of sensors, including voltage, current, frequency and temperature. This module can be deployed at multiple key locations of the electric power grid, including transformer substations, power transmission and distribution line nodes, load centers and access points of important equipment, to achieve comprehensive monitoring of the operation state of the electric power grid.

[0051] First, voltage sensors and current sensors are installed at the high-voltage bus of the transformer substation, the intermediate nodes of the power transmission and distribution line, and the access points of important loads. The voltage sensor can collect the instantaneous value of the node voltage in real time, which is used to monitor the stability of the electric power grid voltage; the current sensor is responsible for capturing the amplitude and direction information of the node current, which is used to evaluate the current distribution and load situation. The sensors send the collected data to the integrated data collection unit of the module through standard communication protocols such as MODBUS or IEC 61850.

[0052] Secondly, temperature sensors are installed on key junctions of the line, the shell of high-load equipment, and the transformer of the power grid to monitor the temperature changes of the equipment. Temperature data is particularly important because temperature rise may indicate that the line is overloaded or that the equipment is abnormal. The sensors transmit data to the central processor of the multi-level perception module through a wireless communication module such as ZigBee or LoRa.

[0053] For frequency detection, a high-precision frequency acquisition device is integrated into the module, installed on the low-voltage side of the main substation, and can monitor the operating frequency of the power grid at a high sampling rate (e.g., 100 times per second) to capture frequency deviations caused by load fluctuations or faults. In addition, the device has anti-interference function and can maintain high-precision measurement in strong electromagnetic interference environment.

[0054] The multi-level perception module 101 also includes a data integration and packaging unit for aggregating, formatting, and packaging data from different types of sensors to form standardized data packets. The unit ensures the consistency of all collected data timestamps through a time synchronization mechanism (e.g., GPS-based clock synchronization), thereby improving the accuracy of data analysis.

[0055] Finally, the module is equipped with reliable data transmission function. The integrated communication module supports multiple communication methods, including wired methods (such as Ethernet or optical fiber) and wireless methods (such as LTE or 5G), to ensure that the collected data can be quickly and stably transmitted to the data processing module 102. To improve the robustness of data transmission, the module also has data buffering and retransmission mechanism, which can temporarily store data when communication is interrupted and resend after recovery, avoiding data loss.

[0056] Through the above design, the multi-level perception module 101 realizes multi-dimensional monitoring of the operating state of the power grid and efficient transmission of data, providing reliable data support for subsequent data processing and fault diagnosis.

[0057] Further, the multi-level perception module is specifically used for:

[0058] Voltage sensors and current sensors are deployed at substations, transmission and distribution lines, and load ends to collect real-time voltage and current data of each node, wherein the voltage sensor is used to detect the instantaneous value of node voltage, and the current sensor is used to monitor the amplitude and direction of current;

[0059] Temperature sensors are installed at high-load equipment and key line locations to regularly detect the temperature rise of equipment shells and lines, and send temperature data to the data processing unit through a wireless communication module;

[0060] A frequency acquisition device is integrated into the main substation to continuously sample changes in the system frequency, capture frequency offsets caused by load fluctuations or faults, and encapsulate all the acquired data before transmitting it to the data processing module.

[0061] The multi-level sensing module fully considers the operation monitoring needs of different locations in the power grid. By deploying various types of sensors in substations, transmission and distribution lines, load terminals and other key areas, it realizes the real-time acquisition and transmission of operation data such as voltage, current, temperature and frequency, providing comprehensive basic data support for monitoring the operation status of the power grid and diagnosing faults.

[0062] In substations, transmission and distribution lines, and at load terminals, voltage and current sensors are strategically deployed to cover key nodes and pathways of the power grid. Voltage sensors are specifically designed to detect instantaneous voltage values ​​at nodes, with accuracy sufficient to meet the high sensitivity requirements for voltage fluctuations during grid operation. These sensors are typically connected to critical points on high-voltage buses and transmission lines, enabling them to detect rapid voltage drops or other abnormal changes during fault occurrences. Current sensors are installed at the inlets of transmission and distribution lines and load equipment, capable of monitoring both current amplitude and direction. By detecting current flow characteristics, these sensors provide crucial data support for fault location and load analysis.

[0063] Temperature sensors are used to monitor temperature rise in high-load equipment and critical wiring locations. The primary goal of temperature monitoring is to prevent overheating problems caused by overload or equipment failure, which could lead to electrical fires or equipment damage. Temperature sensor installation locations are chosen based on the heat-generating parts of the equipment, such as transformer tanks, switchgear enclosures, and high-voltage line connections. Temperature data is sampled at fixed time intervals and transmitted to the data processing unit via a wireless communication module. The wireless communication module supports multiple protocols (such as ZigBee, LoRa, or LTE) and can maintain stable data transmission in complex electromagnetic environments.

[0064] In the main substation, frequency acquisition devices are integrated into the system to continuously sample the system frequency during grid operation. These devices feature high sampling rates and high accuracy, enabling them to capture frequency shifts caused by load fluctuations or faults. For example, when a short circuit or large-scale load switching occurs, the frequency acquisition device can instantly detect changes in the system frequency and transmit this as an initial signal of an anomaly to subsequent data processing modules. To ensure the synchronization of the acquired frequency data, the frequency acquisition device typically incorporates a high-precision clock source, supporting time alignment with data from other sensors.

[0065] All sensor data will be encapsulated by the integrated processing unit of the multi-level perception module after collection, forming data packets with standardized structures. These data packets include voltage, current, temperature, and frequency measurements, timestamps, and device identifiers. The integrated processing unit also has data integrity check functions, which can check the collected data for errors and request retransmission if necessary to ensure data accuracy. After data encapsulation, these data are sent to the data processing module through wired or wireless communication for further analysis and diagnosis.

[0066] Through the above implementation steps, the multi-level perception module can meet the diversified needs of power grid operation state monitoring, and its efficient and reliable monitoring and data transmission functions provide a solid foundation for the operation of the entire power grid real-time fault detection and automatic repair system.

[0067] The data processing module 102 is configured to receive operation data provided by the multi-level perception module, analyze the operation data through embedded intelligent algorithms to identify potential faults, and generate fault diagnosis results when a fault is detected, wherein the fault diagnosis results include fault type, fault location, and impact range.

[0068] The data processing module 102 is one of the core components of the power grid real-time fault detection and automatic repair system. Its function is to receive operation data from the multi-level perception module 101, analyze it using embedded intelligent algorithms, identify potential faults, and generate diagnosis results including fault type, fault location, and impact range. The module design covers the entire process of data reception, preprocessing, analysis, diagnosis, and output, with all operations targeting efficiency and automation.

[0069] When receiving operation data, the data processing module connects with the multi-level perception module through standardized communication interfaces that support multiple communication protocols such as MODBUS, IEC 61850, or TCP / IP-based communication protocols. The module is internally configured with a receiving unit that can handle data streams from multiple sensors simultaneously, including voltage, current, frequency, and temperature parameters. To ensure data integrity, the receiving unit has error correction mechanisms that can detect and repair bit errors during data transmission, and supports temporary data caching to handle short-term communication interruptions.

[0070] After entering the data processing module, the operation data will be standardized and denoised by the preprocessing unit. First, the module normalizes each data stream according to its physical quantity unit to eliminate analysis bias caused by dimensional differences. For example, for voltage and current data, the preprocessing unit will convert them into percentage form based on the rated parameters of the equipment. Second, the module uses sliding window filtering technology and threshold rejection algorithm to clean up high-frequency noise and outliers in the data, ensuring the accuracy of subsequent analysis. At the same time, the preprocessing unit assigns a uniform timestamp to data from different sensors through a time synchronization mechanism to ensure the correctness of time correlation in multi-dimensional data analysis.

[0071] After completing the preprocessing, the module will extract features from the cleaned data, which is the basis for fault diagnosis. The feature extraction unit, combined with the characteristics of power grid operation data, designs a multi-level feature analysis logic. For example, for current data, the module calculates the waveform change rate and the current difference between adjacent nodes, which can effectively reflect the unbalanced condition of power grid load; for voltage data, the module extracts unbalance degree and fluctuation amplitude to identify voltage abnormal areas. In addition, the module also analyzes the dynamic change pattern of frequency data, capturing the change of main frequency offset and harmonic components to provide support for subsequent diagnosis.

[0072] Fault diagnosis is the core function of the data processing module. The module is embedded with a pre-trained intelligent algorithm based on the support vector machine (SVM) model, which is dynamically optimized combined with real-time updated load data. The algorithm receives the extracted feature data as input and generates fault diagnosis results through classification and regression analysis. Specifically, the classification part is used to identify fault types such as short circuit, overload or device anomaly, while the regression analysis calculates the fault location such as specific line or node according to the feature data. To improve the accuracy of diagnosis, the algorithm dynamically adjusts the boundary of the classification model combined with historical sample data and current load distribution, thus adapting to the changes of real-time operation environment.

[0073] After generating the fault diagnosis results, the data processing module will package and output the results, including fault type, fault location and impact range, etc. The module is equipped with a standardized output unit supporting multiple formats of data packaging such as JSON, XML or binary data stream. The output unit can send the diagnosis results to the automatic repair module 103 through high-speed communication interface to trigger repair operations, and at the same time, pass the results to the monitoring and log module 104 for recording and display.

[0074] The data processing module 102 not only can efficiently process massive data in power grid operation, but also can accurately identify and locate faults, providing strong technical support for the stable operation of power grid.

[0075] Further, the intelligent algorithm used by the data processing module is implemented using the following steps:

[0076] The running data collected by the multi-level perception module is decomposed using wavelet transform, and the low-frequency components are extracted by removing high-frequency noise. The low-frequency components are analyzed using sliding window time series analysis to extract characteristic data reflecting the operating state of the power grid, generating power grid characteristic data. The extracted characteristic data includes current waveform change rate, voltage unbalance degree, and frequency offset characteristics.

[0077] The characteristic data is input into a pre-trained support vector machine classifier, and the classification boundary is dynamically optimized based on historical samples and real-time load data to generate a diagnosis result of the fault type. The abnormal operating parameters related to the fault type are marked according to the diagnosis result, and the fault type includes short circuit, overload, line breakage, or equipment abnormality.

[0078] According to the diagnosis result of the fault type and the marked abnormal operating parameters, the physical structure of the power grid is traced back layer by layer, and the source of voltage or current change is analyzed based on the abnormal current propagation path. According to the analysis result of the source of voltage or current change, and combined with the operating characteristics of the equipment and the line, the specific location of the fault is determined, and a fault location result including the identification of the faulty equipment or line is generated.

[0079] Based on the fault location result, combined with the connection relationship of the equipment in the power grid and the real-time load state, the influence range of the fault is analyzed by step-by-step propagation, and the operating state change of the substation, transmission line, and load equipment and the affected area are gradually calculated. When the influence is weakened to a pre-set negligible range, detailed information of the affected area is generated, including equipment identification, operating parameter change value, and specific influence range.

[0080] The intelligent algorithm used in the data processing module realizes comprehensive analysis of the operating state of the power grid, fault diagnosis and positioning, and accurate assessment of the influence range of the fault through a multi-step coordinated analysis and calculation process. The implementation of the algorithm covers data preprocessing, feature extraction, fault diagnosis, physical structure backtracking, and influence range analysis, and the design of each stage ensures its logicality and technical feasibility.

[0081] In the first stage, the algorithm receives raw operational data from the multi-level perception module, including voltage, current, frequency, and temperature signals. These signals are often disturbed by high-frequency noise, so the algorithm first decomposes the data using wavelet transform. Wavelet transform can divide the operational data into different frequency bands, where the high-frequency components contain noise information, and the low-frequency components retain the main power grid operating characteristics. By removing high-frequency components and retaining low-frequency components, the algorithm generates a set of time series data reflecting the dynamic characteristics of the power grid. These data are then processed in segments, and each time series is analyzed using a sliding window technique to capture local changes in operating conditions.

[0082] The results of the sliding window analysis are used for feature extraction, which is the basis of fault diagnosis. The feature extraction stage focuses on three core indicators: current waveform change rate, voltage imbalance, and frequency offset characteristics. The current waveform change rate is obtained by calculating the current change amplitude between adjacent time points, which is used to identify abnormal fluctuations in current; the voltage imbalance is calculated by comparing the voltage differences between different phases of the power grid, reflecting the stability of voltage distribution; the frequency offset characteristic captures the deviation and trend of the power grid operating frequency, which can effectively indicate load fluctuations or fault impacts.

[0083] The extracted feature data is input into a pre-trained support vector machine classifier for fault diagnosis. The support vector machine uses historical sample data for initial training, including various fault types and their corresponding feature distributions, while combining real-time load data to optimize the classification boundary, thereby improving the accuracy and real-time performance of diagnosis. The output of the classifier is the diagnosis result of the fault type, such as short circuit, overload, line breakage, or device abnormality. In addition, the diagnosis result will mark the abnormal operating parameters related to the fault type, such as the current peak value or frequency abnormal value at a specific time point, providing a basis for subsequent analysis.

[0084] Based on the diagnosis result, the algorithm performs a layer-by-layer backtracking analysis on the physical structure of the power grid to determine the source of voltage or current changes. This process uses power grid topology information and abnormal current propagation paths to gradually trace back from affected load devices to power transmission and distribution lines, until the starting node of abnormal changes is found. By combining operating parameters and historical data, the algorithm further analyzes the abnormal behavior of the source node, such as sudden current rise or voltage drop, and finally determines the location of the faulty device or line. The result of this stage is to generate an accurate fault location report including the identification of the faulty device or line.

[0085] After determining the fault location, the algorithm further assesses the overall impact range of the fault on the grid operating state. Through step-by-step propagation analysis, the algorithm simulates the diffusion process of abnormal current or voltage and dynamically calculates its impact range in combination with the operating characteristics of devices and lines. The result of each propagation updates the operating state changes of related devices, including current load increase, voltage drop amplitude increase, or frequency offset diffusion, etc. When the propagation impact weakens to a pre-set negligible range, the algorithm stops propagation calculation and generates a detailed impact report. The report lists the identification of all affected devices, the specific change values of operating parameters, and the geographical or topological distribution of the impact range.

[0086] Through the coordinated work of the above stages, the intelligent algorithm in the data processing module can realize the full-process automatic processing from the preliminary analysis of operating data to fault location and impact assessment, providing technical support for the rapid repair and comprehensive assessment of grid faults.

[0087] Furthermore, the operating data collected by the multi-level perception module are decomposed using wavelet transform, and low-frequency components are extracted by removing high-frequency noise, including:

[0088] According to the following formula (1) and formula (2), the wavelet basis functions ψ j,k (t) and φ j,k (t) are defined as:

[0089]

[0090] Where t represents the sampling time, which is the time coordinate of the grid operating data;

[0091] j represents the scale parameter, which controls the balance between the time resolution and the frequency resolution of the wavelet basis function;

[0092] k represents the time shift parameter, which determines the position of the wavelet on the time axis.

[0093] ψ(t) and φ(t) are the basis functions of the wavelet function, as shown in the following formula (3) and formula (4):

[0094] ψ(t) = exp(-αt 2 )cos(2πβt) (3)

[0095] φ(t) = exp(-αt 2 ) (4)

[0096] Where α is the frequency offset adjustment parameter, which is used to control the time domain extension of the wavelet function, and a small value is usually selected according to the frequency characteristics of the grid signal to enhance the time resolution;

[0097] β is a waveform adjustment parameter, which defines the center frequency of the wavelet function, and is usually selected to be consistent with the main frequency characteristics of the power grid signal;

[0098] t represents the sampling time;

[0099] According to the following formula (5) and (6), the operating data x(t) is multi-scale decomposed to generate high-frequency component D j (t) and low-frequency component A j (t):

[0100]

[0101] D j (t) represents the convolution operation of the operating data x(t) and the wavelet base function ψ j,k (t), which is used to capture the high-frequency change part of the signal.

[0102]

[0103] A j (t) is used to calculate the convolution of the operating data x(t) and the scale function φ j,k (t), which extracts the low-frequency characteristics of the signal. Extract the low-frequency characteristics of the signal.

[0104] According to the following formula (7), the high-frequency component threshold T is calculated:

[0105]

[0106] where η is a proportion factor set based on the statistical characteristics of the power grid operating data. Based on the statistical characteristics of the power grid operating data. Usually, by statistical analysis of historical operating data, a value that can effectively distinguish noise and signal is selected, for example, η can be taken in the range of 1 to 3.

[0107] N is the number of sampling points;

[0108] The high-frequency component |D j (t)|>T is set to zero to remove noise and obtain the denoised high-frequency component;

[0109] By inverse wavelet transform on the denoised high-frequency component and low-frequency component, the low-frequency component reflecting the power grid operating state is reconstructed as the key feature input describing the power grid state.

[0110] Further, the low-frequency component is analyzed by sliding window time series analysis to extract feature data reflecting the power grid operating state, and generate power grid feature data, including:

[0111] The operating data of current, voltage and frequency are respectively wavelet decomposed to generate corresponding low-frequency components Acurrent (t), A voltage (t) and A frequency (t), and a sliding window technique is adopted to analyze each low-frequency component in time series, defining the length W and step S of the sliding window, and extracting various features from different windows as follows.

[0112] Here, the low-frequency component A current (t), A voltage (t) and A frequency (t) can be obtained by the method of the above equations (1)-(7). The sliding window technique effectively captures the local variation characteristics by processing the signal in segments. The length W of the sliding window determines the time span of each segment analysis, which is usually set according to the characteristic period of the power grid signal, for example, when the power grid frequency is 50 Hz, W can be selected in the range of 0.1 seconds to 1 second. The step S defines the interval step number of the window sliding, a smaller step can improve the resolution, while a larger step can reduce the computational burden, and the typical value is 10% to 20% of W.

[0113] According to the following equation (8), the current waveform variation rate R current is calculated as follows:

[0114]

[0115] where N is the number of data segments in the window, N = W / S; m i is the fitting slope of the i-th data segment in the window; m i-1 is the fitting slope of the i-1-th data segment in the window; m i is calculated using the following equation (9):

[0116]

[0117] where W i is the i-th sliding window, whose time range is [t i , t i +W], where W is the window length; t i is the starting time of the i-th sliding window; t is the sampling time point; is the mean value of the time points in the window, which is calculated using the following equation (10):

[0118]

[0119] where n is the number of sampling points in the window; is the mean value of the low-frequency components in the window, which is calculated using the following equation (11):

[0120]

[0121] Wherein, n is the number of sampling points in the window;

[0122] According to the following formula (12), the voltage imbalance degree U is calculated unbalance :

[0123]

[0124] Wherein, t k represents the starting time of the kth sliding window; W is the length of the sliding window; S is the step; the numerator part calculates the voltage difference between adjacent sampling points, and the denominator part uses the maximum and minimum values of the signal to normalize the result, facilitating the comparison of signals with different amplitude ranges.

[0125] According to the following formula (13), the frequency offset characteristic F is calculated offset :

[0126]

[0127] Wherein, ω normal is the normal frequency of the power grid (usually 50Hz or 60Hz); ω inst (t) is the instantaneous frequency at time t, calculated using the following formula (14):

[0128]

[0129] Wherein, H represents the Hilbert transform symbol; arg is the phase angle calculation function;

[0130] The calculated current waveform change rate R current , voltage imbalance degree U unbalance and frequency offset characteristic F offet are integrated to generate power grid feature data.

[0131] Further, the feature data is input into a pre-trained support vector machine classifier, the classification boundary is dynamically optimized based on historical samples and real-time load data, and the diagnosis result of the fault type is generated, including:

[0132] Receive the feature data of the power grid operating state, the feature data including the current waveform change rate, the voltage imbalance degree and the frequency offset characteristic; normalize the input feature data to ensure the consistency of the dimensions of different features;

[0133] Train the support vector machine classifier using a historical sample data set, the historical sample data set including fault type labels and corresponding feature data; the fault types include short circuit, overload, line breakage and equipment abnormality;

[0134] The decision boundary of the classifier is dynamically adjusted in combination with real-time load data, and the classification model is optimized by updating the weighted distance formula between support vectors. The optimization objective is as follows:

[0135]

[0136] where w is the weight vector of the classifier; ξ i is the relaxation variable of the i-th sample; n is the number of samples; β is the load-sensitive weight coefficient; load i is the load data of the i-th sample; load represents the load data of all samples.

[0137] The power grid real-time fault detection and automatic repair system uses a support vector machine classifier to diagnose fault types, combines historical sample data and real-time load data, and generates accurate diagnostic results by dynamically optimizing the classification boundary.

[0138] The input of the support vector machine classifier is the feature data of the power grid operating state, which includes current waveform change rate, voltage unbalance degree, and frequency offset characteristics. Since the dimensions of these feature data are different, for example, the current waveform change rate is expressed in amplitude change, and the frequency offset characteristic is expressed in hertz, normalization processing is required for all feature data before inputting into the classifier. Normalization scales the value of each feature by its maximum value or standard deviation to ensure that the range of all features is the same, usually normalized to the range of [0, 1] or [-1, 1], thereby avoiding the influence of dimensional differences on the performance of the classifier.

[0139] The training process of the support vector machine classifier is based on a pre-prepared historical sample data set. The historical sample data set includes multiple samples of known fault types, each containing corresponding feature data and labels. The labels describe the fault types, such as short circuit, overload, line breakage, or device abnormality. Through these sample data, the classifier can learn the feature patterns corresponding to each fault type. During the training process, the classifier separates different types of fault data by optimizing the classification boundary, so as to accurately classify unknown faults during runtime.

[0140] The optimization objective of the classifier is based on the following formula:

[0141]

[0142] where w represents the weight vector of the classifier, which determines the position and direction of the classification boundary. The classification boundary is a hyperplane in a high-dimensional space that can effectively separate fault types. The first term of the optimization objective controls the complexity of the classification boundary, and a smaller weight value w can prevent overfitting of the classifier.

[0143] The relaxation variable ξ in the second term i is the tolerance for classification errors. For samples that are difficult to classify, the value of ξ i will be larger, allowing the classifier to have some errors on these samples. The weight C determines the weight of the error term, a larger C value tends to strictly reduce the classification error, but may cause overfitting; a smaller C value allows greater classification error, but can improve the generalization ability of the model.

[0144] The formula also contains a load-sensitive weight adjustment term where β is a sensitivity coefficient that controls the degree of influence of real-time load on the adjustment of the classification boundary. This coefficient is usually selected by experimental data optimization, with a value range of 0.1 to 1. The real-time load data load i is the load information of the specific sample in operation, and max(load) is the maximum value of all sample loads for normalization processing. Through the introduction of this term, the classifier can assign higher weights to samples with larger loads, thereby dynamically adjusting the classification boundary to better adapt to the load characteristics in real-time operation environment.

[0145] In actual optimization process, support vector machine uses gradient descent or other optimization algorithms to minimize the above objective function. As the optimization proceeds, the classification boundary will gradually adjust, eventually achieving the effect of maximizing the differentiation of different fault types. After optimization, the classifier outputs the diagnosis results of fault types, including the label of fault type and the confidence score.

[0146] Through the above design, the classifier can dynamically adjust in real time by combining the operation data of the power grid, achieving accurate classification and efficient diagnosis of complex power grid fault types. According to the load characteristics, historical sample data distribution and real-time operation requirements of the actual power grid, parameters C, β and normalization range can be flexibly adjusted to optimize the performance of the classifier and adapt to different application scenarios.

[0147] Further, the diagnosis results of fault types and the labeled abnormal operating parameters are used to backtrack the physical structure of the power grid layer by layer, analyze the source of voltage or current change based on the abnormal current propagation path, including:

[0148] Based on the diagnosis results of fault types and the abnormal operating parameters, the initial node associated with the abnormal data in the physical structure of the power grid is identified; and the real-time operating state of the initial node and the connected upstream and downstream nodes are recorded;

[0149] The physical structure of the power grid is backtracked layer by layer, and the upstream and downstream nodes or lines connected to the initial node along the abnormal current propagation path are checked in sequence, and the propagation direction of current or voltage abnormality is judged by comparing the voltage change difference and current imbalance characteristics of adjacent nodes;

[0150] In each layer of backtracking, historical operation data, topology structure and real-time load state are combined to identify the main path of abnormal propagation, finally determine the initial source of current or voltage change, and mark the initial source as a suspected fault point for subsequent analysis.

[0151] The embodiment analyzes the source of voltage or current change based on the abnormal current propagation path by layer-by-layer backtracking of the physical structure of the power grid, and gradually determines the initial source of the abnormality by combining the diagnosis results of the fault type and the marked abnormal operation parameters, providing key basis for fault location. The specific implementation process is described in detail below.

[0152] First, the system identifies the initial node related to abnormal data in the physical structure of the power grid based on the fault diagnosis results and the marked abnormal operation parameters. The initial node refers to the power grid node or device directly associated with the abnormal operation state in the fault type analysis, such as a specific load end, a substation or a connection point of a transmission and distribution line. In identifying the initial node, the system uses real-time monitoring data (such as current fluctuations, voltage outliers) in combination with the topology structure of the power grid to lock the relevant nodes through parameter matching and logical deduction. For each initial node, the system not only records its abnormal parameter value, but also synchronously extracts the connection relationship of its upstream and downstream nodes, which will be used as the basis for judging the abnormal propagation path in the subsequent backtracking analysis.

[0153] After determining the initial node, the system starts the process of layer-by-layer backtracking of the physical structure of the power grid. The backtracking process starts from the initial node and gradually traces back to the source of the power grid along the propagation path of the abnormal current based on the topology information of the power grid. In each layer of backtracking, the system analyzes the voltage change and current characteristics of the upstream and downstream nodes or lines directly connected to the current node. Specifically, by comparing the voltage change difference between adjacent nodes, the system can determine the propagation direction of the voltage anomaly. At the same time, by analyzing the current imbalance characteristics (such as current amplitude mutation or phase deviation) of adjacent lines, the system further verifies whether the current anomaly continues to the node.

[0154] In the backtracking process, the system also combines historical operation data, real-time load state and topology structure characteristics to optimize the judgment of abnormal propagation path. Historical operation data provides the behavior patterns of nodes or lines under normal and abnormal conditions, such as load fluctuation range, line impedance characteristics, etc. Real-time load state provides dynamic reference for current load distribution and operating conditions, such as whether a node is overloaded or there is power imbalance. By integrating these information, the system can more accurately determine the direction and main path of abnormal propagation, avoiding false judgments caused by fluctuations in a single parameter.

[0155] Finally, the system locks the initial source of the current or voltage change based on the abnormal propagation path. The source node is the main location that triggers the anomaly, which could be an abnormal point caused by equipment failure, short circuit, or other physical problems. When identifying the source node, the system marks it as a suspected fault point and attaches information including its physical location (such as a substation or line number), main anomaly parameter values (such as current peak or voltage drop value), and associated upstream and downstream nodes. These information provides key support for subsequent further analysis and repair operations.

[0156] Through the above implementation steps, the method not only accurately traces the propagation path of current or voltage anomalies, but also accurately identifies the source fault point from complex power grid structures, providing technical support for fast repair and reliable operation of the power grid.

[0157] Furthermore, the specific location of the fault occurrence is determined based on the analysis results of the source of the voltage or current change, combined with the operating characteristics of the equipment and lines, including:

[0158] The operating parameters of the marked suspected fault point and its surrounding equipment are analyzed to check the load current change rate of the equipment, the voltage drop amplitude of the line, and the timing characteristics of current imbalance;

[0159] Combined with the operating characteristics of the equipment and lines, including the rated load capacity of the equipment, the impedance characteristics of the line, and the historical fault mode, the suspected fault point is comprehensively evaluated to determine whether it meets the actual fault characteristics, and the most likely fault point or fault line is selected;

[0160] The final fault point location is confirmed by matching with the power grid topology information.

[0161] Based on the analysis results of the source of the voltage or current change, combined with the operating characteristics of the equipment and lines, the specific location of the fault occurrence is determined, providing an accurate method for power grid fault location. This process gradually confirms the specific location of the fault by comprehensive analysis of the suspected fault point and its surrounding equipment, combined with power grid operating characteristics and topology information, providing accurate basis for subsequent repair operations.

[0162] In the first stage of fault location, the system conducts in-depth analysis of the operating parameters of the marked suspected fault point and its surrounding equipment. These operating parameters include the load current change rate of the equipment, the voltage drop amplitude of the line, and the timing characteristics of current imbalance. The load current change rate can reflect the stability of the equipment operation, for example, a sharp change in current within a short period of time may indicate that the equipment is affected by an anomaly. The voltage drop amplitude of the line is a key indicator of the health of the transmission line, and a significant voltage drop may be caused by an increase in line impedance or a short circuit. In addition, the timing characteristics of current imbalance analysis can reveal whether there are asymmetric loads or phase breaks in a three-phase system, which are usually accompanied by faults.

[0163] In the second stage, the system conducts a comprehensive evaluation of the suspected fault points by combining the operational characteristics of devices and lines. Operational characteristics include the rated load capacity of devices, impedance characteristics of lines, and historical fault patterns. For example, the rated load capacity is used to determine whether the current load of a device exceeds the design range, while the line impedance characteristics can help assess whether the voltage drop amplitude is consistent with normal operation. By introducing historical fault pattern data, the system can identify whether the current situation matches known past fault patterns, further verifying the reliability of suspected fault points. Based on these analysis results, the system prioritizes all suspected fault points, filtering out the most likely fault points or faulty lines.

[0164] Finally, by matching with the topology information of the power grid, the system confirms the final fault point location. Topology information includes the connection relationship of devices and lines in the power grid, as well as the geographical or physical location of each node. The system uses topology information to verify whether the filtered fault points conform to the actual current or voltage propagation path, and checks whether the operation status of upstream and downstream devices or lines is consistent with the fault characteristics. Through this method, false markings caused by abnormal data or accidental fluctuations can be effectively excluded, ensuring the accuracy of the final positioning result.

[0165] The entire fault location process technically realizes multi-level optimization from data analysis to comprehensive judgment and finally to final confirmation. The design of the system ensures that the location of fault points not only depends on a single parameter, but also provides highly accurate fault diagnosis results through multi-dimensional data verification and logical derivation.

[0166] The automatic repair module 103 is used to execute repair operations according to the fault diagnosis results generated by the data processing module, wherein the repair operations include isolating fault areas, reconstructing power grid topology, and restoring normal power supply.

[0167] The core of the automatic repair module 103 is to efficiently and automatically handle power grid faults, and its operation includes three main stages: fault area isolation, topology reconstruction, and power supply recovery.

[0168] Firstly, in the fault area isolation stage, the module initiates isolation control operations using fault diagnosis results received from the data processing module 102. The system maintains a dynamically updated power grid topology database, recording all lines, nodes, and their connection relationships. After receiving the specific fault location, the isolation control unit extracts a list of switch devices directly connected to the fault node or line from the database and establishes control channels with these devices based on real-time communication protocols such as IEC 61850. Isolation instructions are sent to target devices through the channel in a standard format, including the unique identification of the switch, the execution action (such as disconnect), the trigger condition, and the confirmation mechanism. To ensure the reliability of the isolation operation, the module will monitor the state feedback signal of the switch device in real time. If no confirmation signal of successful execution is received within the set time, the module will automatically send repeated instructions or alarm information to the backup control path.

[0169] In the topology reconstruction stage, the module relies on the built-in topology optimization engine, which calculates the optimal power supply path based on the depth search algorithm and load balancing rules. First, the module extracts all node and line information of the non-fault area from the power grid topology database and calculates the remaining transmission capacity of each line combined with real-time load data. Second, the engine treats the power grid as a directed graph, where nodes represent substations or load devices, and edges represent transmission lines. Through dynamic weighting, the edge weight is given as the main parameter of the line's remaining capacity, and the depth-first search or Dijkstra algorithm is used to find the shortest path or optimal path from the power source to the load. The final generated path scheme is sent to the control unit in the form of instructions, including the switch devices that need to be turned on for each line, load scheduling information, and line safety verification parameters.

[0170] The power supply recovery stage is completed by the power supply recovery unit, and its operation process is refined into a step-by-step power supply recovery mechanism. First, the module determines the priority of power supply recovery based on the topology optimization results, usually prioritizing the restoration of lines connected to critical load devices such as hospitals or communication base stations. The specific execution method of the restoration operation is to activate the line in segments and monitor the overall operation state of the power grid after each activation, including voltage fluctuation range, frequency stability, and whether the current distribution exceeds the set threshold. If any abnormalities are detected during the recovery process, the module will immediately suspend subsequent operations and notify the monitoring module to re-evaluate the repair scheme. To prevent load surges from causing power grid instability, the power supply recovery process introduces a load segment activation strategy, such as gradually increasing the load through preset time intervals.

[0171] The automated repair module 103 not only quickly isolates the fault area, dynamically reconstructs the power grid topology, but also gradually restores power supply in a stable and safe manner, significantly improving the reliability and operational efficiency of the power grid.

[0172] Further, the automatic repair module is specifically used for:

[0173] Based on the fault diagnosis results, the circuit breaker control unit is started to isolate the identified fault area, wherein the isolation includes physically disconnecting the faulty line or device from the normally powered part;

[0174] The power grid topology of the non-fault area is calculated, a new power supply path is generated by enabling backup power sources, backup lines or redistributing existing loads, and instructions are sent to the relevant switching devices through the communication module to perform topology reconstruction;

[0175] After confirming the completion of topology adjustment, the voltage level and power supply are adjusted through the transformer to gradually restore power supply to the normal load area.

[0176] The automatic repair module realizes the isolation of power grid fault, dynamic adjustment of topology structure and power supply recovery of non-fault area through a series of coordinated operations, ensuring that the power grid can operate stably and quickly. Based on the fault diagnosis results, the module starts the corresponding repair operation and reduces manual intervention through highly automated processes to improve repair efficiency and accuracy.

[0177] When the data processing module generates fault diagnosis results, the automatic repair module will immediately start the circuit breaker control unit to isolate the fault area. The module first analyzes the specific fault location and impact range in the fault diagnosis results, and matches them with the topology structure database of the power grid to determine the lines or devices that need to be isolated and their associated circuit breaker numbers. The isolation operation is completed by sending instructions to the corresponding circuit breakers or switching devices through the control unit. These instructions include device identification, execution action (such as disconnection) and operation confirmation mechanism. To ensure the reliability of the operation, the module designs a real-time monitoring function to verify the execution of the instructions by receiving the state feedback signals of the circuit breakers. If the target device cannot be successfully disconnected within the specified time, the module will trigger a backup plan, such as resending the isolation instructions or reporting the exception to the monitoring module.

[0178] After completing the isolation of the fault area, the module will enter the topology adjustment stage of the power grid. At this time, the topology optimization engine will take over the operation, which uses the real-time acquired running data of the non-fault area and topology structure information to calculate the new power supply path. Based on the load balancing principle and path priority rule, the optimization engine selects the optimal backup power source, backup line or redistributes the existing load. Specifically, the backup power source can include temporarily enabled generator sets or unused power supply lines, while load redistribution is achieved by dynamically adjusting the transformer output or switching the load connection point. The generated new path scheme will be packaged into a set of instructions and sent to the relevant switching devices through the built-in communication interface of the module. These instructions include the list of devices that need to be connected or switched, load adjustment parameters and safety check rules to ensure the safety and reliability of the topology adjustment execution process.

[0179] Upon completion of topology adjustment, the automated repair module initiates the power restoration process, aiming to gradually restore normal power supply in non-fault areas. Power restoration is handled by the power restoration unit, which operates in a segmented activation manner to avoid secondary problems caused by sudden load increase. Before each activation, the module assesses whether the current load capacity is sufficient to support new power supply requirements by monitoring real-time operating conditions of transformers and lines. If abnormalities such as overload or voltage fluctuations exceeding thresholds are detected, the module suspends the restoration operation and re-evaluates the restoration strategy. During the power restoration process, the module dynamically adjusts transformer voltage levels and output power to accommodate the gradually increasing load. Finally, when power restoration in all load areas is complete, the module notifies the monitoring module to record detailed data of the entire repair process.

[0180] The automated repair module is also equipped with redundant safety mechanisms to quickly switch to backup solutions in case of isolation or topology adjustment failure. To achieve this, the module incorporates real-time anomaly detection and response functions. When any abnormality is detected, the system immediately suspends the current operation and sends the abnormality to the monitoring module, while attempting to re-plan the repair strategy or request human intervention.

[0181] Through the above design, the automated repair module can achieve full-process automation from fault isolation to power restoration, significantly improving the risk resistance of the power grid in terms of flexibility and reliability.

[0182] The monitoring and logging module 104 is used to record the operating state of the power grid, fault diagnosis results and repair operation process, generate historical data reports, and provide fault analysis and operating performance evaluation through the user interface.

[0183] The monitoring and logging module 104 is a key part of the real-time fault detection and automated repair system for the power grid. Its main functions include recording the operating state of the power grid, fault diagnosis results and repair operation process, generating detailed historical data reports, and providing fault analysis and operating performance evaluation through the user interface. This module design combines the dual requirements of real-time and data analysis, providing comprehensive support for the operation and management of the power grid through efficient recording mechanisms, flexible display methods and powerful analysis functions.

[0184] The monitoring and logging module first receives real-time data and operation results through communication interfaces with the data processing module 102 and the automated repair module 103. The received data includes operating state parameters such as voltage, current, frequency and temperature, as well as fault diagnosis types, locations, impact ranges, and repair operation steps and time nodes. The module assigns accurate time information to each record through the timestamp mechanism, ensuring consistency of data from different sources in the time dimension.

[0185] The recording function is the foundation of this module, through the log recording unit to record the key events in the power grid operation in layers. The first layer is the real-time recording layer, which records all the operation data and fault information, ensuring the integrity and traceability of the data. The second layer is the event recording layer, which captures and organizes important events in operation, such as the time points of fault occurrence, repair operation start and end, and warning or abnormal information related to the power grid operation state. The third layer is the report generation layer, which generates formatted historical data reports according to the recorded data, including operation curves, fault summaries and repair operation logs. These reports can be exported in common formats (such as PDF or CSV) for long-term storage and further analysis.

[0186] In order to facilitate users to monitor and interact in real time, the module is equipped with a user interface unit, which presents the operation state and history of the power grid through visualization. The user interface uses a multi-dimensional display method, including real-time monitoring view, event list and performance evaluation dashboard. In the real-time monitoring view, parameters such as voltage, current and frequency are displayed in the form of dynamic charts, and users can intuitively view the current power grid operation state. The event list lists all key events, and users can click on the event to get detailed information, such as the topology of the fault location or the execution steps of the repair operation. The performance evaluation dashboard provides quantitative evaluation of the quality of power grid operation through indicators such as average repair time, fault occurrence frequency and system stability score.

[0187] The monitoring and logging module also has intelligent analysis function, which can conduct deep mining based on historical data. For example, the module can identify common fault patterns through clustering analysis of multiple fault data, and find potential causes related to faults through association analysis. In addition, the module supports simulating past faults and repair processes through the playback function, helping users verify the response speed of the system and the effectiveness of the repair strategy.

[0188] The design of this module fully considers the data security and reliability. All log records are stored redundantly in local and cloud servers to prevent data loss. The module also uses a permission management mechanism to ensure that only authorized users can access sensitive data or perform data export operations.

[0189] The monitoring and logging module 104 realizes the whole process function from data recording to advanced analysis, providing solid support for the transparent management and continuous optimization of power grid operation state.

[0190] Further, the monitoring and logging module is specifically used for:

[0191] Real-time operation data, including voltage, current, frequency, and temperature parameters, are acquired by the state acquisition unit. When a fault occurs, abnormal data are recorded, and snapshots of the operation state before and after the fault are generated.

[0192] During the fault repair process, the specific steps and execution times of isolation operations, topology adjustments, and power restoration are recorded in real time.

[0193] After repair is complete, a complete log report is generated, including grid operation state data, fault cause analysis, and repair process, which is displayed to operators through a visual interface and stored.

[0194] The monitoring and log module is an important component of the power grid real-time fault detection and automatic repair system, responsible for comprehensive monitoring of power grid operation state, accurate recording of fault data, and complete tracking of repair process, providing reliable data support for subsequent operation evaluation and optimization. The module works collaboratively through the state acquisition unit, real-time recording function, and log report generation unit to ensure transparency of operation state and systematic management of information.

[0195] The module acquires real-time operation data of the power grid through the state acquisition unit, including voltage, current, frequency, and temperature. Data collection is based on communication interfaces with the multi-level perception module. The acquisition unit can receive data streams at high frequency and attach accurate timestamps to each data to ensure time sequence consistency. In normal operation state, the module continuously stores collected data and monitors parameter changes in real time. If an anomaly is detected, such as significant deviation of voltage or frequency, the module will immediately trigger abnormal data recording. Abnormal recording includes not only abnormal parameter values but also captures key operation states before and after the fault, generating snapshots of operation state. These snapshots provide preliminary inference basis for fault causes by analyzing current and historical data.

[0196] During the fault repair process, the monitoring and log module plays a core recording function. The module tracks every step of the automatic repair module, including isolation operations of circuit breakers, topology adjustments, and specific steps of power restoration. At each operation execution, the module records relevant time points, identification information of execution devices, and operation results, such as whether the isolation circuit breaker successfully disconnected or whether there are load abnormalities in power restoration. In addition, the module captures dynamic changes of power grid operation parameters during the repair process to assess the effectiveness of repair measures in subsequent analysis. For example, during the activation process of segmented power restoration, the module can record current and voltage fluctuations when each segment of the line is restored to determine whether the power supply strategy needs to be optimized.

[0197] After the fault repair is completed, the monitoring and logging module will enter the log report generation phase. The module will organize and summarize all data from the entire repair process to generate a complete log report. This report includes a detailed record of the power grid's operating status, the initial parameters of the fault and their changing trends, the complete repair operation process, and the final recovery result. In addition, the report includes a detailed analysis of the fault's causes, such as identifying possible fault sources through correlation analysis or predicting potential risks of similar faults through trend analysis. The log report is stored in a standardized format, supporting multiple file types such as PDF and CSV, facilitating long-term archiving or export for use by other systems.

[0198] To facilitate rapid decision-making and long-term management by operators, the module is equipped with a visual interface. This interface uses an intuitive graphical display, presenting operational data and the repair process through graphs, topology diagrams, and timelines. Operators can use the interface to view the power grid status in real time, browse key fault information, or replay the repair process. Furthermore, the interface supports customized performance evaluation functions, such as calculating the average repair time or the frequency of fault occurrence by analyzing log reports, helping power grid managers optimize system operation strategies.

[0199] The monitoring and logging module is designed with data security and reliability in mind. It employs a dual-redundancy storage mechanism, synchronously saving data to both local and cloud servers to prevent data loss due to hardware failure or network interruptions. The module also includes access control to ensure that only authorized users can view or export sensitive data.

[0200] Through the above design, the monitoring and logging module realizes full-process management from data acquisition to fault analysis, from real-time monitoring to historical archiving, providing a solid technical guarantee for the efficient operation of the power grid.

[0201] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

Claims

1. A real-time fault detection and automatic restoration system for power grid, characterized in that, The utility model relates to an electric power grid fault diagnosis and repair system, comprising: a multi-level perception module for collecting operating data of the power grid through multiple types of sensors and sending the collected operating data to a data processing module; wherein the operating data includes voltage, current, frequency and temperature; a data processing module for receiving the operating data provided by the multi-level perception module; analyzing the operating data through embedded intelligent algorithms to identify potential faults and generating fault diagnosis results when a fault is detected, wherein the fault diagnosis results include fault type, fault location and impact range; an automated repair module for performing repair operations according to the fault diagnosis results generated by the data processing module, wherein the repair operations include isolating the fault area, restructuring the power grid topology and restoring normal power supply; a monitoring and logging module for recording the operating state of the power grid, fault diagnosis results and repair operation process, generating historical data reports, and providing fault analysis and operating performance evaluation through a user interface; wherein the intelligent algorithm used by the data processing module is implemented using the following steps: using wavelet transform to decompose the operating data collected by the multi-level perception module, extracting low-frequency components by removing high-frequency noise; performing sliding window time series analysis on the low-frequency components to extract feature data reflecting the operating state of the power grid, generating power grid feature data; wherein the extracted feature data includes current waveform change rate, voltage imbalance and frequency offset characteristics; inputting the feature data into a pre-trained support vector machine classifier, dynamically optimizing the classification boundary based on historical samples and real-time load data, and generating fault type diagnosis results; according to the diagnosis results, marking abnormal operating parameters related to the fault type, including short circuit, overload, line break or equipment anomaly; based on the fault type diagnosis results and the marked abnormal operating parameters, analyzing the source of voltage or current change by layer-by-layer backtracking the physical structure of the power grid based on the abnormal current propagation path; determining the specific location of the fault occurrence based on the analysis results of the source of voltage or current change, combined with the operating characteristics of the equipment and lines, and generating fault positioning results including fault equipment or line identification; based on the fault positioning results, combining the connection relationship of the equipment in the power grid and the real-time load state, analyzing the impact range of the fault through step-by-step propagation, and gradually calculating the operating state change of the substation, transmission line and load equipment and the affected area; when the impact is weakened to a pre-set negligible range, generate detailed information of the affected area, including equipment identification, operating parameter change value and specific impact range; wherein, based on the fault type diagnosis results and the marked abnormal operating parameters, analyzing the source of voltage or current change by layer-by-layer backtracking the physical structure of the power grid based on the abnormal current propagation path, comprising: based on the fault type diagnosis results and the abnormal operating parameters, identifying the initial node associated with abnormal data in the physical structure of the power grid; at the same time, recording the real-time operating state of the initial node and the connected upstream and downstream nodes; The physical structure of the power grid is traced back layer by layer, and the upstream and downstream nodes or lines connected to the initial node along the abnormal current propagation path are checked in sequence, and the propagation direction of the current or voltage anomaly is determined by comparing the voltage change difference and current imbalance characteristics of adjacent nodes. In each layer of backtracking, the historical operation data, topological structure and real-time load state are combined to identify the main path of abnormal propagation, and finally determine the initial source of current or voltage change, and mark the initial source as a suspected fault point.

2. The power grid real-time fault detection and automatic restoration system of claim 1, wherein, The multi-level perception module is specifically used for: Voltage sensors and current sensors are deployed at substations, transmission and distribution lines, and load ends to collect real-time voltage and current data of each node. The voltage sensor is used to detect the instantaneous value of the node voltage, and the current sensor is used to monitor the amplitude and direction of the current. Temperature sensors are installed at high-load devices and key line locations to detect the temperature rise of device housings and lines at regular intervals. The temperature data is transmitted to the data processing unit through the wireless communication module. A frequency acquisition device is integrated into the main substation to continuously sample the system frequency changes, capture the frequency deviation caused by load fluctuations or faults, and transmit all collected data to the data processing module after encapsulation.

3. The power grid real-time fault detection and automatic restoration system of claim 1, wherein, The automatic repair module is specifically used for: Based on the fault diagnosis result, the circuit breaker control unit is started to isolate the identified fault area, wherein the isolation includes physically disconnecting the fault line or device from the normal power supply part; The power grid topological structure of the non-fault area is calculated, a new power supply path is generated by enabling the standby power supply, standby line or redistributing the existing load, and the communication module sends instructions to the related switching devices to perform topological reconstruction; After confirming the completion of topological adjustment, the voltage level and power supply power are adjusted through the transformer to gradually restore the power supply of the normal load area.

4. The power grid real-time fault detection and automatic restoration system of claim 1, wherein, The monitoring and logging module is specifically used for: Real-time operation data, including voltage, current, frequency and temperature parameters, are obtained through the state acquisition unit. Abnormal data are recorded when a fault occurs, and a snapshot of the running state before and after the fault is generated. During the fault repair process, the specific steps and execution time of the isolation operation, topological adjustment and power supply recovery are recorded in real time. After the repair is completed, a complete log report including the power grid operation state data, fault cause analysis and repair process is generated, which is displayed to the operator through the visual interface and stored.

5. The power grid real-time fault detection and automatic restoration system of claim 1, wherein, The running data collected by the multi-level perception module are decomposed using wavelet transform, and the low-frequency components are extracted by removing high-frequency noise, including: The wavelet basis function is defined according to the following equation (1) and equation (2) and : ； ； wherein, denotes a sampling time; denotes a scale parameter; denotes a time shift parameter; and are basis functions of the wavelet function as shown in the following equation (3) and equation (4): ； ； wherein, is a frequency offset adjustment parameter; is a waveform adjustment parameter; denotes a sampling time; The operation data is decomposed into high frequency components and low frequency components according to the following equations (5) and (6) :​​ ； ； The high frequency component threshold is calculated according to the following equation (7) : ； wherein, is a scaling factor set based on statistical characteristics of grid operation data; is a number of sampling points; Will The high-frequency components are set to zero to remove noise, thus obtaining the denoised high-frequency components. The low-frequency components reflecting the power grid operation state are reconstructed by inverse wavelet transform on the denoised high-frequency components and low-frequency components, which are used as key features to describe the power grid state.

6. The grid real-time fault detection and automatic restoration system of claim 5, wherein, The low-frequency components are analyzed by sliding window time series analysis to extract feature data reflecting the power grid operation state, and power grid feature data are generated, including: The running data of current, voltage and frequency are respectively wavelet decomposed to generate corresponding low-frequency components , and , and the time series analysis is performed on each low-frequency component by using the sliding window technology, the sliding window length and the step are defined, and the following features are extracted from different windows: The rate of change of the current waveform is calculated according to the following equation (8) : ； wherein, is the number of data segments within the window, ; is the slope of the fit for the th data segment within the window; is the slope of the fit for the th data segment within the window; is calculated using the following equation (9): ； wherein, is the start time of the th sliding window, whose time range is wherein, is the window length; is the start time of the th sliding window; is the sampling time point; is the mean of the time points within the window, calculated using the following equation (10): ； wherein is the number of sampling points within the window; is the mean of the low frequency components within the window, calculated using equation (11) as follows: ； wherein, is the number of sampling points within the window; The voltage unbalance degree is calculated according to the following equation (12) : ； wherein, denotes the start time of the th sliding window; is the sliding window length; is the step size; The frequency offset characteristic is calculated according to the following equation (13) : ； wherein is the normal grid frequency; is the time of the instantaneous frequency, calculated using the following equation (14): ； wherein denotes the Hilbert transform symbol; is a phase angle calculation function; The calculated current waveform change rate , voltage imbalance , and frequency offset characteristics are integrated to generate grid feature data.

7. The grid real-time fault detection and automatic restoration system of claim 1, wherein, The feature data are input into the pre-trained support vector machine classifier, the classification boundary is dynamically optimized based on historical samples and real-time load data, and the diagnosis result of the fault type is generated, including: Receive feature data of power grid operating state, which includes current waveform change rate, voltage imbalance and frequency offset characteristics; normalize the input feature data to ensure the consistency of different features; Train the support vector machine classifier using a historical sample data set, which includes fault type labels and corresponding feature data; fault types include short circuit, overload, line breakage and equipment abnormality; Dynamically adjust the decision boundary of the classifier combined with real-time load data, optimize the classification model by updating the weighted distance formula between support vectors, and the optimization objective is as follows: ； wherein, is a weight vector of the classifier; is a relaxation variable for the i-th sample; is a number of samples; is a load sensitive weight coefficient; is load data for the i-th sample; denotes load data for all samples.

8. The grid real-time fault detection and automatic restoration system of claim 1, wherein, According to the analysis result of the source of voltage or current change, and combined with the operating characteristics of equipment and lines, the specific location of fault occurrence is determined, including: Analyze the operating parameters of the marked suspected fault point and its surrounding equipment, check the load current change rate of the equipment, the voltage drop amplitude of the line, and the time sequence characteristics of the current imbalance; Combined with the operating characteristics of equipment and lines, including the rated load capacity of equipment, the impedance characteristics of lines and historical fault patterns, comprehensively evaluate whether the suspected fault point meets the actual fault characteristics, and select the most possible fault point or fault line; Confirm the final fault point location by matching with the power grid topology information.

Images