Power grid fault early warning system based on machine learning, and related apparatus

Abstract

The present invention relates to the technical field of data processing, and disclosed are a power grid fault early warning system based on machine learning, and a related apparatus. The system comprises: an initial monitoring module that monitors the voltage, current and load of a power grid to identify and mark an anomalous data point, so as to obtain an anomalous pattern identification result; a load fluctuation analysis module that positions an anomalous node on the basis of the anomalous pattern identification result, so as to obtain anomalous node positioning information; a trend analysis and prediction module that evaluates and predicts a fault node and time on the basis of the anomalous node positioning information, so as to obtain a risk prediction result; a strategy formulation and response module that formulates an operation instruction on the basis of the risk prediction result, so as to obtain a fault prevention measure; a fault simulation and test module that uses the fault prevention measure to perform virtual fault simulation, and performs iterative optimization on a strategy, so as to obtain a test evaluation result; and a fault early warning and optimization module that optimizes a fault prevention strategy on the basis of the test evaluation result, so as to obtain a more accurate target fault early warning and a faster and more effective target response strategy.

Description

Machine Learning-Based Power Grid Fault Early Warning System and Related Devices Technical Field

[0001] This invention relates to the fields of data processing and fault prediction technology, and in particular to a power grid fault early warning system and related devices based on machine learning. Background Technology

[0002] Power grid fault early warning systems are a key technology applied in the power industry. They aim to predict and warn of impending faults by monitoring power grid operation and environmental variables. This system utilizes various sensors to collect real-time power grid data, such as voltage, current, and frequency, as well as environmental factors like temperature and humidity. The collected data is then fed into analytical models to identify abnormal patterns leading to faults and to issue timely warnings. This allows operators to take necessary measures to avoid or mitigate the impact of faults, thereby ensuring the stable and safe operation of the power grid.

[0003] Current technologies for predicting and managing power grid faults primarily rely on traditional fault indicators and early warning systems. These systems lack in-depth analysis and real-time response capabilities for complex data fluctuations. Especially with the increasing scale of power grids and the growing complexity of data types, the limitations of traditional technologies in handling multi-node and multi-variable data lead to inaccurate fault predictions and an inability to effectively anticipate and handle sudden and complex fault modes. Furthermore, existing technologies employ a reactive maintenance strategy, addressing faults only after they occur. This strategy not only increases maintenance costs but also impacts the stable operation of the power grid and the reliability of power supply. In emergency situations, this low-accuracy fault prediction model and delayed response model may lead to wider service disruptions and security risks. Therefore, improving the fault prediction accuracy and fault response speed of power grid fault early warning systems has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides a power grid fault early warning system and related devices based on machine learning, which can improve the accuracy of power grid fault prediction and the speed of fault response, thereby solving the technical problems of low accuracy of fault prediction mode and relatively slow response mode in the prior art.

[0005] Firstly, a machine learning-based power grid fault early warning system is provided. This system includes an initial monitoring module, a load fluctuation analysis module, a trend analysis and prediction module, a strategy formulation and response module, a fault simulation and testing module, and a fault early warning and optimization module.

[0006] The initial monitoring module identifies and marks abnormal data points by monitoring the voltage, current and load of the power grid, and obtains the abnormal mode identification results;

[0007] The load fluctuation analysis module locates abnormal nodes based on the abnormal mode identification results, and obtains the abnormal node location information.

[0008] The trend analysis and prediction module assesses and predicts the fault nodes and times based on the abnormal node location information, and obtains risk prediction results.

[0009] The strategy formulation and response module formulates operational instructions based on risk prediction results, thereby obtaining preventive measures against failures.

[0010] The fault simulation and testing module uses preventative measures to simulate virtual faults and iteratively optimizes the strategies to obtain test evaluation results.

[0011] The fault warning and optimization module optimizes the fault prevention strategy based on the test evaluation results, and obtains the target fault warning and target response strategy.

[0012] Secondly, a power grid fault early warning device based on machine learning is provided, wherein...

[0013] The first processing module is used to identify and mark abnormal data points by monitoring the voltage, current and load of the power grid, and obtain the abnormal pattern identification result;

[0014] The second processing module is used to locate abnormal nodes based on the abnormal pattern identification results and obtain abnormal node location information.

[0015] The third processing module is used to evaluate and predict the fault node and time based on the abnormal node location information, and obtain the risk prediction result.

[0016] The fourth processing module is used to formulate operation instructions based on the risk prediction results and obtain fault prevention measures;

[0017] The fifth processing module is used to simulate virtual faults using preventative measures, iteratively optimize the strategy, and obtain test evaluation results.

[0018] The sixth processing module is used to optimize the fault prevention strategy based on the test evaluation results, and obtain the target fault warning and target response strategy.

[0019] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the processing steps of the aforementioned machine learning-based power grid fault early warning system.

[0020] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the processing steps of the aforementioned machine learning-based power grid fault early warning system.

[0021] In the aforementioned scheme implemented by the machine learning-based power grid fault early warning system and related devices, the initial monitoring module monitors the voltage, current, and load of the power grid, identifies and marks abnormal data points, and obtains abnormal pattern marking results. This allows the load fluctuation analysis module to locate abnormal nodes based on the abnormal pattern marking results, obtaining abnormal node location information. Furthermore, the trend analysis and prediction module evaluates and predicts fault nodes and times based on the abnormal node location information, obtaining risk prediction results. This enables the strategy formulation and response module to formulate operation instructions based on the risk prediction results, obtaining fault prevention measures. The fault simulation and testing module can use the fault prevention measures to conduct virtual fault simulation and iteratively optimize the strategy, obtaining test evaluation results. Finally, the fault early warning and optimization module can optimize the fault prevention strategy based on the test evaluation results, obtaining accurate target fault early warning and fast and effective target response strategies. This effectively improves the accuracy of power grid fault prediction and the speed of fault response, contributing to achieving more accurate and efficient power grid fault early warning. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 is a schematic diagram of an application environment of a power grid fault early warning system based on machine learning in one embodiment of the present invention;

[0024] Figure 2 is a schematic diagram of a processing flow of a power grid fault early warning system based on machine learning in one embodiment of the present invention;

[0025] Figure 3 is a schematic diagram of a power grid fault early warning device based on machine learning in one embodiment of the present invention;

[0026] Figure 4 is a structural schematic diagram of a computer device according to an embodiment of the present invention;

[0027] Figure 5 is another structural schematic diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] The machine learning-based power grid fault early warning system provided in this invention can be applied in the application environment shown in Figure 1. In this environment, the client communicates with the machine learning-based power grid fault early warning system via a network. As shown in Figure 1, the machine learning-based power grid fault early warning system may include an initial monitoring module, a load fluctuation analysis module, a trend analysis and prediction module, a strategy formulation and response module, a fault simulation and testing module, and a fault early warning and optimization module.

[0030] For example, taking a power grid fault early warning scenario, managers can receive analytical data or prediction results from a machine learning-based power grid fault early warning system via a client. Correspondingly, the initial monitoring module in the machine learning-based power grid fault early warning system can monitor the voltage, current, and load of the power grid to identify and mark abnormal data points, obtaining abnormal pattern identification results. This allows the load fluctuation analysis module to locate abnormal nodes based on the abnormal pattern identification results, obtaining abnormal node location information. Furthermore, the trend analysis and prediction module can evaluate and predict the fault node and time based on the abnormal node location information, obtaining risk prediction results. This enables the strategy formulation and response module to formulate operational instructions based on the risk prediction results, obtaining fault prevention measures. The fault simulation and testing module can then use these preventative measures to conduct virtual fault simulations and iteratively optimize the strategy, obtaining test evaluation results. Finally, the fault early warning and optimization module can optimize the fault prevention strategy based on the test evaluation results, obtaining accurate target fault early warning and a rapid and effective target response strategy, and feeding this target fault early warning and target response strategy back to the client. Correspondingly, the client can receive target fault warnings and target response strategy feedback from the server, and can display these warnings and feedback on the client for target users to query or browse. By adopting the machine learning-based power grid fault warning system provided in this application, the accuracy of power grid fault prediction and the speed of fault response can be effectively improved, which is conducive to achieving the goal of more accurate and efficient power grid fault warning.

[0031] The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers. The invention will now be described in detail through specific embodiments.

[0032] Please refer to Figure 2, which is a schematic diagram of a processing flow of a machine learning-based power grid fault early warning system provided in an embodiment of the present invention, including the following steps:

[0033] S10: The initial monitoring module identifies and marks abnormal data points by monitoring the voltage, current and load of the power grid, and obtains the abnormal mode identification result.

[0034] The initial monitoring module monitors the voltage, current, and load of the power grid to identify data points that deviate from normal behavior patterns. It then analyzes the power grid data by setting thresholds and identifies anomalies in these data points to obtain anomaly pattern identification results. These anomaly pattern identification results include, but are not limited to, anomaly identification time, anomaly type, and anomaly intensity.

[0035] It is important to understand that the initial monitoring module, by monitoring the voltage, current, and load of the power grid, identifies and marks abnormal data points to obtain anomaly pattern identification results. This refers to the process of obtaining the anomaly pattern identification results. Specifically, step S10, where the initial monitoring module identifies and marks abnormal data points by monitoring the voltage, current, and load of the power grid to obtain anomaly pattern identification results, may include the following steps:

[0036] S11: The voltage and current monitoring submodule samples the voltage, current and load of the power grid to obtain sampled data;

[0037] S12: The voltage and current monitoring submodule performs frequency analysis and stability testing on the sampled data and generates a stability evaluation result;

[0038] S13: The load analysis submodule uses the stability assessment results to dynamically track the power grid load and obtain load data;

[0039] S14: The load analysis submodule performs periodic analysis on the load data to obtain load consistency analysis results;

[0040] S15: The anomaly pattern recognition submodule uses the load consistency analysis results to analyze anomaly patterns and obtain anomaly data;

[0041] S16: The abnormal pattern recognition submodule performs data pattern comparison on the abnormal data to obtain the abnormal pattern identification result.

[0042] Among them, the voltage and current monitoring submodule can continuously sample and record signals by monitoring the voltage, current and load of the power grid, and perform frequency analysis and stability testing on the sampled data. For example, by adjusting the sampling interval and analysis cycle, the monitoring process can be optimized to further generate stability assessment results.

[0043] Monitoring the voltage, current, and load of the power grid can be achieved through precise measurement of voltage and current values ​​per second, and calculation of power consumption per unit time, through continuous signal sampling and recording. The sampling frequency can be set to 1000Hz to capture instantaneous changes, and noise is removed through filtering to ensure data clarity and accuracy. For frequency analysis, Fourier transform can be used to convert the time-domain signal to the frequency-domain signal to analyze the main frequency components. Stability detection can be performed by calculating the standard deviation and peak factor of voltage and current to assess the amplitude of power grid load fluctuations. The monitoring process can be optimized by adjusting the sampling interval and analysis period to generate stability assessment results.

[0044] Optionally, the formula for calculating stability testing can be as follows:

[0045] Where S represents the stability test result; N represents the total number of samples; i represents the index of a single sample; x i This represents the voltage or current value of the i-th sample, where μ represents the average value; σ represents the standard deviation; and k represents the voltage or current value of the i-th sample. a This indicates that adjusting the parameters can enhance sensitivity to extreme values. By increasing the value of k, the ability to determine grid instability can be improved.

[0046] The load analysis submodule can dynamically track the power grid load using the stability assessment results, perform periodic analysis on the load data, further adjust the threshold to identify load fluctuations, and perform anomaly detection by comparing real-time data with the set threshold, thereby obtaining the load consistency analysis results.

[0047] Based on the stability assessment results, dynamic tracking of the power grid load can be performed, recording the highest and lowest load values ​​every minute. The mean and variance of the data are calculated using a sliding window to determine the normal fluctuation range of the load. Periodic analysis of the load data is conducted, and the autocorrelation function is used to identify the periodic characteristics of load changes. A threshold for identifying load fluctuations is adjusted; the threshold is based on a comparison of the load data from the same time the previous day with the current data. Anomaly detection is performed by comparing real-time data with the set threshold, obtaining load consistency analysis results.

[0048] Optionally, the following formula can be used to calculate the load deviation:

[0049] Where D represents the calculated load deviation; L t L represents the load value at the current time point t. avg σ represents the average load value at the same time point on the previous day. L This indicates the standard deviation of the load.

[0050] The anomaly pattern identification submodule can use the load consistency analysis results to analyze anomaly patterns, such as performing time series analysis on the identified anomaly data to adjust identification parameters and match differentiated power grid states, and further identify abnormal behavior through data pattern comparison to obtain anomaly pattern identification results.

[0051] The results of load consistency analysis can be used to analyze abnormal patterns. First, an abnormal pattern is defined as an event in which a sudden load change exceeds a set threshold. Time series analysis is performed on the identified abnormal data, and trend analysis is conducted using moving average and exponential smoothing techniques to identify short-term abnormal changes. The identification parameters are adjusted and matched to different power grid conditions. The parameters include window size and smoothing coefficient to adapt to different monitoring needs. Abnormal behavior is identified by comparing data patterns, thus obtaining the abnormal pattern identification results.

[0052] S20: The load fluctuation analysis module locates the abnormal nodes based on the abnormal mode identification results and obtains the abnormal node location information.

[0053] The load fluctuation analysis module can monitor the current and voltage of multiple nodes in the power grid based on the abnormal pattern identification results, and record fluctuations exceeding safety standards, thereby locating abnormal nodes and obtaining abnormal node location information. This abnormal node location information includes, but is not limited to: node identifier, fluctuation duration, and fluctuation frequency.

[0054] It is important to understand that the load fluctuation analysis module, by locating the abnormal node based on the abnormal pattern identification result and obtaining the abnormal node location information, refers to the process of obtaining the abnormal node location information. Specifically, step S20, where the load fluctuation analysis module locates the abnormal node based on the abnormal pattern identification result and obtains the abnormal node location information, may include the following steps:

[0055] S21: The real-time monitoring submodule monitors the operation status of the power grid in real time based on the abnormal mode identification result and obtains a real-time data snapshot;

[0056] S22: The fluctuation detection submodule extracts current and voltage data from the real-time data snapshot, identifies abnormal fluctuations, and generates excessive fluctuation analysis results;

[0057] S23: The node positioning submodule uses the results of the out-of-standard fluctuation analysis to identify and locate nodes with abnormal fluctuations and establish abnormal node positioning information.

[0058] The real-time monitoring submodule can collect current and voltage data from multiple nodes of the power grid based on the abnormal mode identification results, and monitor the operation of the power grid in real time by setting the sampling frequency and data recording interval, thereby obtaining real-time data snapshots.

[0059] Based on the anomaly pattern identification results, current and voltage data can be collected from multiple nodes of the power grid. First, the sampling frequency for current and voltage is set to 100 times per second, with a data recording interval of 1 minute. By adding data buffering, the integrity and accuracy of the data are maintained under high-frequency acquisition conditions. Real-time data processing algorithms are used to perform preliminary filtering and noise reduction on the collected current and voltage data. A sliding window technique is used to calculate the average current and voltage within each time window, generating a time-series data snapshot. This enables real-time monitoring of the power grid's operating status, thereby obtaining real-time data snapshots.

[0060] Alternatively, the formula for data filtering can be as follows:

[0061] Where Filtered represents the filtered value; i represents the number of samples; k represents the index of the sampling point; n represents the number of sampling points in the sliding window; x k w represents the data from the k-th sample. k This represents the weighting coefficient, which can be reduced by a decay factor to give higher weight to recent data.

[0062] The fluctuation detection submodule can further extract current and voltage data from the real-time data snapshot, and monitor data fluctuations by setting thresholds and marking data that exceeds the thresholds. This allows for analysis of data change trends and fluctuation frequencies to identify abnormal fluctuations and generate out-of-standard fluctuation analysis results.

[0063] To extract current and voltage data from real-time data snapshots, the fluctuation thresholds for current and voltage can be set to 596 standard deviations and 3% standard deviation, respectively. These thresholds can be set based on historical data distribution. Statistical analysis methods are then used to calculate the degree of deviation for each data point, marking data points with deviations exceeding the thresholds. In this way, each measurement of current and voltage can be monitored in real time, data exceeding the thresholds can be marked, and the trend and frequency of data changes can be analyzed using Fast Fourier Transform to identify abnormal fluctuations and generate out-of-range fluctuation analysis results.

[0064] Optional, the formula for fluctuation monitoring can be as follows:

[0065] Where Fluctuation represents the result of fluctuation monitoring; m represents the total number of measurements; i represents the index of the number of measurements; x i Let represent the i-th measurement of current or voltage, and μ represent the average value of the measurement sequence. This formula can determine the degree of fluctuation by calculating the absolute deviation of all measurements from the average value.

[0066] The node location submodule can use the above-mentioned excessive fluctuation analysis results to analyze the distribution of abnormal data points and locate nodes in the power grid. By comparing data patterns and abnormal behaviors, it can identify and locate nodes with abnormal fluctuations, thereby establishing abnormal node location information.

[0067] The distribution of anomalous data points can be analyzed using the results of out-of-range fluctuation analysis. First, all data points marked as anomalous are collected. Then, cluster analysis is used to classify the data points by their geographical and grid topological locations. By analyzing the central tendency of data points in each category, nodes in the power grid can be located. Finally, by comparing historical data patterns with currently observed anomalous behavior, nodes with abnormal fluctuations can be effectively identified and located, establishing anomalous node location information.

[0068] S30: The trend analysis and prediction module evaluates and predicts the fault node and time based on the abnormal node location information to obtain the risk prediction result.

[0069] The trend analysis and prediction module can assess changes in node data using the abnormal node location information and predict fault nodes and times by drawing potential fault maps, thereby obtaining risk prediction results. These risk prediction results include, but are not limited to: risk level, potential impact area, and impact duration.

[0070] It is important to understand that the trend analysis and prediction module, which evaluates and predicts the fault node and time based on the abnormal node location information to obtain the risk prediction result, refers to the process of obtaining the risk prediction result. Specifically, step S30, where the trend analysis and prediction module evaluates and predicts the fault node and time based on the abnormal node location information to obtain the risk prediction result, may include the following steps:

[0071] S31: The conventional data comparison submodule obtains the current abnormal node and historical fluctuation data based on the abnormal node location information;

[0072] S32: The conventional data comparison submodule uses time series analysis to identify the consistency and key deviations of the pattern based on the current abnormal node and the historical fluctuation data, and obtains similar data and abnormal data;

[0073] S33: The conventional data comparison submodule constructs and obtains data comparison analysis results based on the similarity data and the abnormal data;

[0074] S34: The fault trend analysis submodule clusters the data points based on the data comparison and analysis results to generate a fault trend map;

[0075] S35: The risk prediction submodule, based on the fault trend map, uses a time series prediction algorithm to predict the fault nodes and times in the future time period, and obtains a reference fault prediction result;

[0076] S36: The risk prediction submodule performs risk assessment and prediction model optimization by applying trend analysis and behavioral pattern recognition based on the reference fault prediction results, and obtains risk prediction results.

[0077] The conventional data comparison submodule can compare the current abnormal node with the previous fluctuation data based on the abnormal node location information, and use time series analysis to identify pattern consistency and key deviations. This can include data synchronization and deviation measurement, and further evaluate similarity and abnormality to construct the data comparison analysis results.

[0078] Based on the location information of abnormal nodes, real-time data and stored historical fluctuation data of the current abnormal node can be extracted, and data preprocessing can be performed, including noise removal and normalization. Secondly, time series analysis methods can be used to compare data points point by point through sliding window technology. The mean squared error (MSE) of the current data and historical data is calculated for each window to evaluate the pattern consistency and key deviations at each time point. The deviation threshold method can be used to determine the degree of anomaly. Each deviation value is compared with a preset threshold to construct data comparison analysis results and evaluate similarity and anomaly. This can include data synchronization and deviation measurement.

[0079] Optionally, the calculation formula for the data comparison and analysis results can be as follows:

[0080] Where MSE represents the data comparison analysis result; n represents the total number of data points in the window; i represents the data point index; Xa i Represents real-time data points. This represents the historical data point corresponding to the real-time data point.

[0081] The fault trend analysis submodule can cluster data points by drawing fluctuation maps and trend lines based on the data comparison and analysis results, in order to reveal the development path and pattern of potential faults, including map updates and fault prediction model adjustments, thereby generating a fault trend map.

[0082] Based on the results of data comparison and analysis, by comparing failure modes in historical data with current data, kernel density estimation can be applied to perform cluster analysis of data points. By identifying high-density areas in the dataset, potential failure development paths can be revealed. Historical trend data can also be compared with current trend data to update the map to reflect the latest operating status, and fluctuation maps and trend lines can be drawn. Visual tools can be used to intuitively display data clustering results and potential failure modes, including map updates and failure prediction model adjustments, and generate failure trend maps.

[0083] The risk prediction submodule can use time series prediction algorithms based on the fault trend map to predict the fault nodes and times in the future time period, and perform risk assessment and prediction model optimization by applying trend analysis and behavioral pattern recognition to establish and obtain risk prediction results.

[0084] Based on the fault trend map, by integrating the fault trend analysis results, the weighted moving average method can be used to predict the fault nodes and times in the future time period, analyze the behavioral patterns in the historical data of each node, calculate the future risk level according to the development trend of the pattern, conduct risk assessment by applying trend analysis and behavioral pattern recognition, comprehensively consider various influencing factors, optimize the prediction model, and establish risk prediction results.

[0085] Optionally, the risk prediction submodule can use the following formula to predict the fault nodes and times in the future time period based on the fault trend map and using a time series prediction algorithm to obtain a reference fault prediction result:

[0086] Among them, X t The value at time t in the time series represents the reference fault prediction result; c represents the constant term; α represents the trend term T. t-1 Adjustment factor; T t-1 This represents the trend term from the previous period; p represents the order of the autoregressive term; i represents the order index of the autoregressive term; φ i X represents the coefficient of the autoregressive term; t-i β represents the value at time ti in the time series; i θ represents the adjustment coefficient of the autoregressive term; q represents the order of the moving average term; j represents the order index of the moving average term; θ j The coefficients of the moving average term; ∈ t-j γ represents the random error term at time tj; j The coefficient representing the moving average term is δ; the seasonal adjustment parameter is δ; D t Indicator variables representing seasonal factors; ∈ tThis represents the random error term at time t.

[0087] The trend term T of the previous period can be calculated. t-1 By introducing it as an independent variable into the model, the sensitivity to time series trends is improved, and an autoregressive term φ is introduced. i X t-i and moving average term θ j ∈ t-j And introduce an adjustment coefficient β for the term. i and γ j To dynamically adjust the influence of historical data on current forecasts, a seasonal adjustment parameter δ is added, and the result is multiplied by the seasonal indicator variable D. t To account for the impact of seasonal factors on the time series, the model prediction result X is calculated. t The specific steps for weighting coefficients include parameter optimization using historical data, such as determining the optimal β by minimizing the sum of squared prediction errors. i γ j By combining the above steps with δ, a comprehensive risk prediction model is formed.

[0088] S40: The strategy formulation and response module formulates operation instructions based on the risk prediction results to obtain fault prevention measures.

[0089] The strategy formulation and response module can formulate operational instructions based on the risk prediction results. These instructions may include operations such as adjusting load, switching lines, and reconfiguring power grid resources to avoid potential faults and thus obtain fault prevention measures. These fault prevention measures include, but are not limited to, emergency response levels, resource allocation details, and key operational points.

[0090] It should be understood that the strategy formulation and response module, based on the risk prediction results, formulates operational instructions to obtain fault prevention measures, referring to the process of obtaining fault prevention measures. Specifically, step S40, where the strategy formulation and response module formulates operational instructions based on the risk prediction results to obtain fault prevention measures, may include the following steps:

[0091] S41: The operation instruction formulation submodule adjusts the power grid load allocation and formulates emergency switch operation procedures based on the risk prediction results, and constructs an operation adjustment plan;

[0092] S42: The resource adjustment submodule optimizes the allocation of power grid resources based on the operation adjustment scheme to obtain a reference configuration result;

[0093] S43: The resource adjustment submodule adjusts and matches the predicted fault scenarios according to the reference configuration results, and generates resource configuration results;

[0094] S44: The fault prevention submodule uses the resource configuration results to optimize the power grid operating parameters and upgrade the early warning system to obtain fault prevention measures.

[0095] The operation instruction formulation submodule can refine the operation instructions based on the risk prediction results, analyze the load demand, adjust the power grid load allocation according to the predicted risks, and formulate emergency switch operation procedures, thereby constructing an operation adjustment plan.

[0096] Based on the risk prediction results, information on high-risk areas and time periods can be obtained from the data center first. Historical and predicted power load data in specific areas can be analyzed. The optimal load allocation strategy for the power grid in different time periods can be calculated based on the data. A linear programming model can be used to adjust the load allocation to balance overall energy efficiency and security. By focusing on load adjustments during periods with high predicted risk, specific operation instructions can be designed, such as starting backup power and adjusting the load of the main power grid. Emergency switch operation procedures can also be developed, including instructions for switching between manual and automatic modes. Finally, an operation adjustment plan can be constructed.

[0097] Optional, the formula for load sharing adjustment. It can be as follows: L d =k b ×(P req -P sup )

[0098] Among them, L d Indicates the adjusted load distribution; P req P represents the projected total load demand; sup Indicates current supply capacity; k b This indicates a coefficient adjusted based on risk assessment to ensure sufficient power support can be provided during high-risk periods.

[0099] The resource adjustment submodule can optimize the allocation of power grid resources based on the operation adjustment scheme, such as line switching and reallocation of key resources, and analyze the resource utilization efficiency under the current operating environment to adjust and match the predicted fault scenarios, thereby generating resource allocation results.

[0100] Based on the operational adjustment plan, the current power grid resource allocation can be assessed and optimized. By using a resource allocation model, power lines in the power grid can be dynamically switched, and the allocation of key resources, such as the reallocation of transformers and distribution stations, can be optimized to cope with predicted high-risk events. The optimal resource allocation plan can be calculated, and resources can be adjusted to match the predicted fault scenarios based on the current operating environment and historical fault data, generating resource allocation results.

[0101] Optionally, the formula for calculating resource optimization allocation can be as follows: R c=f(P cur ,P opt )

[0102] Among them, R c Indicates the result of resource allocation; P cur Indicates the current resource status; P opt represents the resource configuration state after optimization based on fault prediction; f represents the resource configuration function, used to consider the difference between the current state and the optimization objective.

[0103] The fault prevention submodule can utilize the resource configuration results to further mitigate anticipated faults through various measures, such as optimizing power grid operating parameters and strengthening supervision, including regular inspections and early warning upgrades.

[0104] By utilizing resource allocation results and integrating optimized resource allocation with real-time monitoring data, power grid operating parameters, such as voltage and frequency regulation, can be adjusted periodically to maintain network stability and efficiency. Enhanced monitoring measures can be implemented, including upgrading fault detection and early warning systems, and conducting regular inspections of lines and equipment. These measures aim to minimize and avoid anticipated faults, thus achieving fault prevention.

[0105] Optionally, the formula for fault prevention regulation can be as follows: F p =γ c ×O p

[0106] Among them, F p Indicates the effectiveness of fault prevention; O p Indicates the optimized operating parameters; γ c This represents the prevention effectiveness coefficient adjusted based on historical failure data, and can be used to indicate the ability to enhance failure prevention capabilities through optimized operating parameters.

[0107] S50: The fault simulation and testing module uses the fault prevention measures to perform virtual fault simulation and iteratively optimizes the strategy to obtain test evaluation results.

[0108] The fault simulation and testing module can use the aforementioned fault prevention measures to perform virtual fault simulations, and adjust operating parameters and iteratively optimize the strategy through simulation feedback to obtain test evaluation results. These test evaluation results include, but are not limited to, simulation accuracy, prediction deviation, and operating response time.

[0109] It is important to understand that the fault simulation and testing module, which uses the aforementioned fault prevention measures to perform virtual fault simulation and iteratively optimizes the strategy to obtain the test evaluation result, refers to the process of obtaining the test evaluation result. Specifically, step S50, where the fault simulation and testing module uses the aforementioned fault prevention measures to perform virtual fault simulation and iteratively optimizes the strategy to obtain the test evaluation result, may include the following steps:

[0110] S51: The parameter adjustment submodule performs virtual fault simulation based on the fault prevention measures to obtain simulation feedback optimization parameters;

[0111] S52: The parameter adjustment submodule constructs an operation parameter adjustment table based on the simulation feedback optimization parameters;

[0112] S53: The strategy iteration submodule uses the operation parameter adjustment table to optimize the fault response strategy and obtain the strategy optimization result;

[0113] S54: The strategy optimization and evaluation submodule uses the strategy optimization results to adjust the test parameters using a Bayesian optimization algorithm to obtain the test evaluation results.

[0114] The parameter adjustment submodule can perform virtual fault simulation based on the aforementioned fault prevention measures to adjust operating parameters, such as the fault response speed and fault detection threshold. This allows for parameter optimization based on simulation feedback, and the operation parameter adjustment table can be constructed with reference to response time and accuracy.

[0115] Virtual fault simulations can be performed based on preventative measures. A high-fidelity simulation platform generates virtual fault events, and the fault response speed and detection threshold are tested in a control environment. Based on the simulation results, the relationship between response time and fault detection accuracy is analyzed, and operating parameters are adjusted to reduce response latency and improve fault detection accuracy. After each simulation, operating parameters can be gradually adjusted based on the differences between actual and expected results to optimize fault response and detection thresholds, generating a detailed operating parameter adjustment table. This table lists the optimal parameter settings under different fault types and conditions.

[0116] Optionally, the formula for adjusting the operating parameters can be as follows:

[0117] Among them, P adj Indicates the adjusted operating parameter value; T resp,i T represents the response time of the i-th simulation; acc,i N represents the detection accuracy corresponding to the i-th simulation; c Indicates the number of simulations; i represents the index of the number of simulations.

[0118] The strategy iteration submodule can use the data in the operation parameter adjustment table to optimize the fault response strategy, such as adjusting the fault isolation and recovery order, and adjust the strategy details according to the results of each iteration to optimize the accuracy of decision-making and response time, so as to obtain the strategy optimization result.

[0119] Data from the operation parameter adjustment table can be used to optimize the fault response strategy. By simulating different fault conditions, the fault isolation and recovery sequence can be adjusted to minimize system downtime and resource waste. Based on the results of each iteration, the module adjusts the strategy in detail, gradually improving the accuracy of decision-making and response time to achieve the optimal fault handling process. Through repeated iterations, the strategy optimization results are obtained, ensuring that faults can be responded to and handled quickly and effectively in practical applications.

[0120] Optionally, the formula for calculating the strategy optimization can be as follows:

[0121] Among them, S opt τ represents the result of the optimized strategy. j This represents the total time for fault handling in the j-th iteration; J represents the iteration number, and j represents the iteration number index.

[0122] The strategy optimization and evaluation submodule can use the strategy optimization results, such as by using a Bayesian optimization algorithm, to perform multiple rounds of testing and evaluation. Each round of testing can be based on the evaluation results of the previous round to adjust the test parameters, thereby summarizing the test data and establishing and obtaining the test evaluation results.

[0123] The results can be optimized through a series of test evaluation processes to verify and refine the fault handling strategy. By building a learning model based on the results of previous iterations of the test, multiple rounds of testing are performed. In each round, the test parameters are adjusted based on the evaluation results of the previous round to refine the strategy and improve the effectiveness of fault handling. In this way, test evaluation results are established to help determine the most effective fault response and resource allocation strategy.

[0124] Optionally, the strategy optimization and evaluation submodule can implement the steps of adjusting the test parameters using a Bayesian optimization algorithm based on the strategy optimization results to obtain the test evaluation results using the following formula: μ(x)=k(x,θ) T (K(θ)+σ n 2 I+λL) -1 y

[0125] Where μ(x) represents the predicted mean at position x, i.e., the test evaluation result; k(x,θ) represents the covariance vector between x adjusted according to parameter θ and the training data points; k(x,θ)T This represents the transpose of k(x,θ); K(θ) represents the covariance matrix between training data points adjusted according to the parameter θ; σ n λ represents the noise term; I represents the identity matrix; λ represents the regularization coefficient; L represents the Laplacian matrix; and y represents the observed values ​​of the objective function at the training data points.

[0126] Optionally, the parameter θ is optimized by maximizing the marginal likelihood function to ensure that the kernel function accurately simulates the behavior of the objective function. The parameter λ is determined through cross-validation to balance the effects of fitting and regularization. The Laplacian matrix L is constructed based on the spatial relationships of the training data points, reflecting the geographical or functional similarity between the data points. These parameters work together on the kernel matrix K and the covariance vector k(x,θ). The predicted objective function value for location x is calculated using the updated prediction model. By combining the above steps, the optimal parameter combination is obtained.

[0127] S60: The fault warning and optimization module optimizes the fault prevention strategy based on the test evaluation results to obtain the target fault warning and target response strategy.

[0128] The fault warning and optimization module can optimize fault prevention strategies based on the test and evaluation results, such as updating the training materials for the operation team and updating the operation team in real time via the network, thereby obtaining the adjusted fault warning and response strategy, i.e., the target fault warning and target response strategy. The adjusted fault warning and response strategy includes, but is not limited to: warning activation time, key response team, and warning communication path.

[0129] It should be understood that the fault warning and optimization module, based on the test evaluation results, optimizes the fault prevention strategy to obtain the target fault warning and target response strategy. This refers to the process of obtaining the target fault warning and target response strategy. Specifically, step S60, where the fault warning and optimization module optimizes the fault prevention strategy based on the test evaluation results to obtain the target fault warning and target response strategy, may include the following steps:

[0130] S61: The fault detection submodule analyzes the operation log based on the test evaluation results to obtain the abnormal frequency and abnormal operation;

[0131] S62: The fault detection submodule classifies and marks the anomalies according to the anomaly frequency and the anomaly operation, and obtains an anomaly classification and marking table;

[0132] S63: The education update submodule uses the anomaly classification label table to conduct training on fault response and obtain an updated training material set;

[0133] S64: The strategy update submodule adjusts and optimizes the fault warning and response based on the updated training material set to obtain a reference fault warning and reference response strategy.

[0134] S65: The strategy update submodule performs a simulation exercise based on the reference fault warning and the reference response strategy, and verifies the implementation effect of the reference response strategy to obtain the target fault warning and the target response strategy.

[0135] The fault detection submodule can perform operation log analysis based on the test evaluation results to identify operational anomalies with abnormal frequency and impact, and classify and mark the anomalies in combination with performance monitoring data, such as recording the performance and frequency of each anomaly, thereby constructing an anomaly classification and marking table.

[0136] Based on test evaluation results, operation logs can be extracted and combined with performance monitoring data to perform frequency analysis on operation anomalies in the logs. Automated tools can be used to identify anomaly patterns, distinguishing between normal operation fluctuations and true operation anomalies. For each type of anomaly identified, the module records its manifestation, frequency, and impact in detail. Machine learning technology is used to classify anomalies, automatically labeling each type of anomaly and generating an anomaly classification label table. This anomaly classification label table can include a detailed description of the anomaly and its frequency of occurrence, providing a reference for subsequent processing.

[0137] Optionally, the formula for calculating anomaly classification can be as follows: C e =f(D log D perf )

[0138] Among them, C e Indicates the anomaly type after classification; D log Indicates operation log data; D perf represents performance monitoring data; f represents a classification function, which can be used to convert logs and performance data into anomaly classification tags.

[0139] The education update submodule can use the anomaly classification and labeling table to recompile the training materials and operating guidelines for the operations team, and conduct case analysis on the identified anomalies to implement training on fault response, thereby obtaining an updated set of training materials.

[0140] An anomaly classification and labeling table can be used to recompile training materials and operating guidelines for the operations team based on the labeled anomaly types. Case studies of each anomaly can be analyzed to develop targeted training plans, thereby enhancing the team's understanding and response speed to faults. Through the analysis of actual cases and simulation of fault responses, training materials can be updated module by module to ensure that the team can quickly identify and effectively handle various anomalies in actual operations, ultimately resulting in an updated set of training materials.

[0141] The strategy update submodule can adjust and optimize the fault warning and response based on the updated training material set, and organize simulation exercises through the internal network to verify the implementation effect of the strategy in real time, thereby obtaining the adjusted fault warning and response strategy, namely the target fault warning and target response strategy.

[0142] Based on the updated training materials, the fault warning and response strategies can be adjusted and optimized. New fault response strategies can be tested by organizing simulation exercises, such as collecting exercise data in real time through the internal network to evaluate the effectiveness and implementation of the strategies. The collected data can be used to make minor adjustments to the strategies to ensure that various fault situations can be warned and responded to efficiently and accurately in actual operation, thereby establishing the adjusted fault warning and response strategies.

[0143] Optionally, the formula for calculating the strategy adjustment can be as follows: S new =g(S old ,R sim )

[0144] Among them, S new Indicates the adjusted strategy; S old Indicates the original strategy; R sim represents the result data of the simulation exercise; g represents the policy adjustment function, which can be used to optimize the original policy based on the exercise results.

[0145] This application achieves accurate identification and threshold analysis of data point anomalies by monitoring grid voltage, current, and load and comparing them with historical behavior data. This improves the accuracy and response speed of fault prediction. Based on real-time monitoring of multi-node current and voltage, it can effectively detect and record data fluctuations that exceed safety standards. The strategy significantly improves the efficiency of fault location. By comparing the location information of abnormal nodes with historical data, it can assess the trend of future faults and predict fault nodes and times through statistical tools, thereby enhancing the comprehensiveness and foresight of the prediction.

[0146] Based on risk prediction results, operational instructions such as load adjustment, line switching, and grid resource allocation can effectively avoid the occurrence of predicted faults and greatly improve the stability and security of grid operation. Through iterative optimization of fault simulation and testing, as well as adjustments to fault early warning and response strategies, the adaptability and operational accuracy of the maintenance team to strategy execution can be enhanced, thereby ensuring efficient and dynamic updates of grid fault management.

[0147] As can be seen, in the above scheme, the initial monitoring module identifies and marks abnormal data points by monitoring the voltage, current, and load of the power grid, obtaining abnormal pattern identification results. This allows the load fluctuation analysis module to locate abnormal nodes based on the abnormal pattern identification results, obtaining abnormal node location information. Furthermore, the trend analysis and prediction module evaluates and predicts fault nodes and times based on the abnormal node location information, obtaining risk prediction results. Consequently, the strategy formulation and response module can formulate operation instructions based on the risk prediction results, obtaining fault prevention measures. This allows the fault simulation and testing module to use the fault prevention measures to perform virtual fault simulation and iteratively optimize the strategy, obtaining test evaluation results. Finally, the fault early warning and optimization module can optimize the fault prevention strategy based on the test evaluation results, obtaining accurate target fault early warning and fast and effective target response strategies. This effectively improves the accuracy of power grid fault prediction and the speed of fault response, contributing to achieving more accurate and efficient power grid fault early warning.

[0148] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0149] In one embodiment, a machine learning-based power grid fault early warning device is provided, which corresponds one-to-one with the machine learning-based power grid fault early warning system described in the previous embodiment. As shown in Figure 3, the machine learning-based power grid fault early warning device includes a first processing module 101, a second processing module 102, a third processing module 103, a fourth processing module 104, a fifth processing module 105, and a sixth processing module 106. Detailed descriptions of each functional module are as follows:

[0150] The first processing module 101 is used to identify and mark abnormal data points by monitoring the voltage, current and load of the power grid, and obtain abnormal mode identification results;

[0151] The second processing module 102 is used to locate the abnormal node according to the abnormal pattern identification result and obtain the abnormal node location information.

[0152] The third processing module 103 is used to evaluate and predict the fault node and time based on the abnormal node location information to obtain the risk prediction result.

[0153] The fourth processing module 104 is used to formulate operation instructions based on the risk prediction results to obtain fault prevention measures;

[0154] The fifth processing module 105 is used to perform virtual fault simulation using the aforementioned fault prevention measures, and to iteratively optimize the strategy to obtain test evaluation results.

[0155] The sixth processing module 106 is used to optimize the fault prevention strategy based on the test evaluation results, and obtain the target fault warning and target response strategy.

[0156] In one embodiment, the first processing module 101 is used to identify and mark abnormal data points by monitoring the voltage, current, and load of the power grid, and to obtain anomaly pattern identification results, specifically for:

[0157] The voltage, current, and load of the power grid are sampled to obtain sampled data;

[0158] Frequency analysis and stability testing are performed on the sampled data to generate stability assessment results;

[0159] Based on the stability assessment results, dynamic tracking of the power grid load is performed to obtain load data;

[0160] Periodic analysis is performed on the load data to obtain load consistency analysis results;

[0161] Using the load consistency analysis results, abnormal patterns are analyzed to obtain abnormal data;

[0162] The abnormal data is compared with data patterns to obtain abnormal pattern identification results.

[0163] In one embodiment, the second processing module 102 is used to locate the abnormal node based on the abnormal pattern identification result to obtain abnormal node location information, specifically for:

[0164] Based on the abnormal mode identification results, the operation status of the power grid is monitored in real time to obtain real-time data snapshots;

[0165] Extract current and voltage data from the real-time data snapshot, identify abnormal fluctuations, and generate analysis results of excessive fluctuations;

[0166] Using the results of the above-mentioned fluctuation analysis, nodes with abnormal fluctuations are identified and located, and abnormal node location information is established.

[0167] In one embodiment, the third processing module 103 is used to evaluate and predict the fault node and time based on the abnormal node location information to obtain a risk prediction result, specifically for:

[0168] Based on the abnormal node location information, obtain the current abnormal node and historical fluctuation data;

[0169] Based on the current abnormal node and the historical fluctuation data, time series analysis is used to identify the consistency and key deviations of the pattern, and similarity data and abnormal data are obtained.

[0170] Based on the similarity data and the abnormal data, a data comparison analysis result is constructed and obtained;

[0171] Based on the data comparison and analysis results, the data points are clustered to generate a fault trend map.

[0172] Based on the fault trend map, a time series prediction algorithm is used to predict the fault nodes and times in the future time period, and a reference fault prediction result is obtained.

[0173] Based on the reference fault prediction results, trend analysis and behavioral pattern recognition are applied to perform risk assessment and prediction model optimization to obtain risk prediction results.

[0174] In one embodiment, the fourth processing module 104 is used to formulate operation instructions based on the risk prediction results to obtain fault prevention measures, specifically for:

[0175] Based on the risk prediction results, the power grid load allocation is adjusted and an emergency switch operation procedure is formulated to construct an operation adjustment plan.

[0176] Based on the aforementioned operational adjustment scheme, the power grid resources are optimized and a reference configuration result is obtained.

[0177] Based on the reference configuration results, adjust and match the predicted fault scenarios to generate resource configuration results;

[0178] The fault prevention submodule utilizes the resource configuration results to optimize power grid operating parameters and upgrade early warning systems, thereby obtaining fault prevention measures.

[0179] In one embodiment, the fifth processing module 105 is used to perform virtual fault simulation using the aforementioned fault prevention measures, and to iteratively optimize the strategy to obtain test evaluation results. Specifically, it is used for:

[0180] Based on the aforementioned fault prevention measures, a virtual fault simulation is performed to obtain simulation feedback optimization parameters;

[0181] Based on the simulation feedback optimization parameters, construct an operation parameter adjustment table;

[0182] The fault response strategy is optimized using the aforementioned operating parameter adjustment table to obtain the strategy optimization result;

[0183] Based on the optimization results of the aforementioned strategy, the test parameters are adjusted using a Bayesian optimization algorithm to obtain the test evaluation results.

[0184] In one embodiment, the fifth processing module 105 is specifically used for:

[0185] The following formula is used to implement the steps of adjusting the test parameters using a Bayesian optimization algorithm based on the optimization results of the stated strategy, and obtaining the test evaluation results: μ(x)=k(x,θ) T (K(θ)+σ n 2 I+λL) -1 y

[0186] Where μ(x) represents the predicted mean at position x, i.e., the test evaluation result; k(x,θ) represents the covariance vector between x adjusted according to parameter θ and the training data points; k(x,θ) T This represents the transpose of k(x,θ); K(θ) represents the covariance matrix between training data points adjusted according to the parameter θ; σ n λ represents the noise term; I represents the identity matrix; λ represents the regularization coefficient; L represents the Laplacian matrix; and y represents the observed values ​​of the objective function at the training data points.

[0187] In one embodiment, the sixth processing module 106 is used to optimize the fault prevention strategy based on the test evaluation results to obtain the target fault early warning and target response strategy, specifically for:

[0188] Based on the test evaluation results, operation log analysis is performed to obtain the frequency of anomalies and abnormal operations.

[0189] Based on the abnormal frequency and the abnormal operation, the abnormalities are classified and marked to obtain an abnormal classification and marking table;

[0190] Using the aforementioned anomaly classification label table, training on fault response is conducted to obtain an updated training material set;

[0191] Based on the updated training material set, the fault warning and response are adjusted and optimized to obtain a reference fault warning and reference response strategy;

[0192] Based on the reference fault warning and the reference response strategy, a simulation exercise is conducted to verify the implementation effect of the reference response strategy, thereby obtaining the target fault warning and the target response strategy.

[0193] This invention provides a power grid fault early warning device based on machine learning. By monitoring the voltage, current, and load of the power grid, it identifies and marks abnormal data points to obtain abnormal pattern identification results. Based on the abnormal pattern identification results, abnormal nodes can be located to obtain abnormal node location information. Furthermore, based on the abnormal node location information, the fault node and time can be evaluated and predicted to obtain risk prediction results. Based on the risk prediction results, operation instructions can be formulated to obtain fault prevention measures. The fault prevention measures are then used to simulate virtual faults, and the strategy is iteratively optimized to obtain test evaluation results. Based on the test evaluation results, the fault prevention strategy can be optimized to obtain accurate target fault early warning and fast and effective target response strategies. This can effectively improve the accuracy of power grid fault prediction and the speed of fault response, which is conducive to achieving more accurate and efficient power grid fault early warning.

[0194] Specific limitations regarding machine learning-based power grid fault early warning devices can be found in the above description of limitations for machine learning-based power grid fault early warning systems, and will not be repeated here. Each module in the aforementioned machine learning-based power grid fault early warning device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0195] In one embodiment, a computer device is provided, which can be a server, and its internal structure diagram is shown in Figure 4. The computer device includes a processor, memory, a network interface, and a database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a machine learning-based power grid fault early warning system server-side.

[0196] In one embodiment, a computer device is provided, which can be a client, and its internal structure diagram is shown in Figure 5. The computer device includes a processor, memory, network interface, display screen, and input device connected via a system bus. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface of the computer device is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a machine learning-based power grid fault early warning system client-side.

[0197] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:

[0198] By monitoring the voltage, current, and load of the power grid, abnormal data points are identified and marked to obtain abnormal pattern identification results;

[0199] The abnormal node is located based on the abnormal pattern identification result to obtain the abnormal node location information;

[0200] Based on the abnormal node location information, the fault node and time are evaluated and predicted to obtain the risk prediction result;

[0201] Based on the risk prediction results, operational instructions are formulated to obtain fault prevention measures;

[0202] The aforementioned fault prevention measures are used to simulate virtual faults, and the strategy is iteratively optimized to obtain test evaluation results.

[0203] Based on the test and evaluation results, the fault prevention strategy is optimized to obtain the target fault early warning and target response strategy.

[0204] This invention provides a computer device that monitors the voltage, current, and load of a power grid, identifies and marks abnormal data points, and obtains anomaly pattern identification results. Based on these results, abnormal nodes can be located to obtain anomaly node location information. Furthermore, based on this location information, fault nodes and their timing can be evaluated and predicted to obtain risk prediction results. Operational instructions can then be formulated based on these risk prediction results to obtain fault prevention measures. These measures can be used to simulate virtual faults, and the strategies can be iteratively optimized to obtain test and evaluation results. Based on these results, fault prevention strategies can be optimized to obtain accurate target fault warnings and rapid, effective target response strategies. This effectively improves the accuracy of power grid fault prediction and the speed of fault response, facilitating more accurate and efficient power grid fault warnings.

[0205] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0206] By monitoring the voltage, current, and load of the power grid, abnormal data points are identified and marked to obtain abnormal pattern identification results;

[0207] The abnormal node is located based on the abnormal pattern identification result to obtain the abnormal node location information;

[0208] Based on the abnormal node location information, the fault node and time are evaluated and predicted to obtain the risk prediction result;

[0209] Based on the risk prediction results, operational instructions are formulated to obtain fault prevention measures;

[0210] The aforementioned fault prevention measures are used to simulate virtual faults, and the strategy is iteratively optimized to obtain test evaluation results.

[0211] Based on the test and evaluation results, the fault prevention strategy is optimized to obtain the target fault early warning and target response strategy.

[0212] This invention provides a computer-readable storage medium that monitors the voltage, current, and load of a power grid, identifies and marks abnormal data points, and obtains anomaly pattern identification results. Based on these results, abnormal nodes can be located to obtain anomaly node location information. Furthermore, based on this location information, fault nodes and their timing can be evaluated and predicted to obtain risk prediction results. Operational instructions can then be formulated based on these risk prediction results to obtain fault prevention measures. These measures can be used to simulate virtual faults, and the strategies can be iteratively optimized to obtain test and evaluation results. Based on these results, fault prevention strategies can be optimized to obtain accurate target fault warnings and rapid, effective target response strategies. This effectively improves the accuracy of power grid fault prediction and the speed of fault response, facilitating more accurate and efficient power grid fault warnings.

[0213] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0214] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0215] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0216] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A power grid fault early warning system based on machine learning, characterized in that, The machine learning-based power grid fault early warning system includes an initial monitoring module, a load fluctuation analysis module, a trend analysis and prediction module, a strategy formulation and response module, a fault simulation and testing module, and a fault early warning and optimization module. The initial monitoring module identifies and marks abnormal data points by monitoring the voltage, current and load of the power grid, and obtains abnormal pattern identification results; The load fluctuation analysis module locates abnormal nodes based on the abnormal mode identification results, and obtains abnormal node location information. The trend analysis and prediction module evaluates and predicts the fault node and time based on the abnormal node location information to obtain the risk prediction result. The strategy formulation and response module formulates operation instructions based on the risk prediction results to obtain fault prevention measures. The fault simulation and testing module uses the aforementioned fault prevention measures to perform virtual fault simulation and iteratively optimizes the strategy to obtain test evaluation results. The fault warning and optimization module optimizes the fault prevention strategy based on the test evaluation results to obtain the target fault warning and target response strategy.

2. The power grid fault early warning system based on machine learning according to claim 1, characterized in that, The initial monitoring module includes a voltage and current monitoring submodule, a load analysis submodule, and an anomaly pattern recognition submodule, wherein... The voltage and current monitoring submodule samples the voltage, current and load of the power grid to obtain sampled data; The voltage and current monitoring submodule performs frequency analysis and stability testing on the sampled data and generates a stability assessment result. The load analysis submodule uses the stability assessment results to dynamically track the power grid load and obtain load data. The load analysis submodule performs periodic analysis on the load data to obtain load consistency analysis results; The anomaly pattern recognition submodule uses the load consistency analysis results to analyze anomaly patterns and obtain anomaly data; The abnormal pattern recognition submodule performs data pattern comparison on the abnormal data to obtain the abnormal pattern identification result.

3. The power grid fault early warning system based on machine learning according to claim 2, characterized in that, The load fluctuation analysis module includes a real-time monitoring submodule, a fluctuation detection submodule, and a node location submodule, wherein... The real-time monitoring submodule monitors the power grid's operating status in real time based on the abnormal mode identification results, and obtains real-time data snapshots. The fluctuation detection submodule extracts current and voltage data from the real-time data snapshot, identifies abnormal fluctuations, and generates excessive fluctuation analysis results. The node positioning submodule uses the results of the out-of-range fluctuation analysis to identify and locate nodes with abnormal fluctuations, and establish abnormal node positioning information.

4. The power grid fault early warning system based on machine learning according to claim 3, characterized in that, The trend analysis and prediction module includes a conventional data comparison submodule, a fault trend analysis submodule, and a risk prediction submodule, wherein... The regular data comparison submodule obtains the current abnormal node and historical fluctuation data based on the abnormal node location information; The conventional data comparison submodule uses time series analysis to identify the consistency and key deviations of the pattern based on the current abnormal node and the historical fluctuation data, and obtains similar data and abnormal data. The conventional data comparison submodule constructs and obtains data comparison analysis results based on the similarity data and the abnormal data; The fault trend analysis submodule clusters the data points based on the data comparison and analysis results to generate a fault trend map. The risk prediction submodule uses the fault trend map and a time series prediction algorithm to predict the fault nodes and times in the future time period, and obtains reference fault prediction results. The risk prediction submodule performs risk assessment and prediction model optimization based on the reference fault prediction results, using trend analysis and behavioral pattern recognition, to obtain risk prediction results.

5. The power grid fault early warning system based on machine learning according to claim 4, characterized in that, The strategy formulation and response module includes an operation instruction formulation submodule, a resource adjustment submodule, and a fault prevention submodule. Based on the risk prediction results, the operation instruction formulation submodule adjusts the power grid load allocation and formulates emergency switch operation procedures, and constructs an operation adjustment plan. The resource adjustment submodule optimizes the allocation of power grid resources based on the operation adjustment scheme to obtain a reference configuration result; The resource adjustment submodule adjusts and matches the predicted fault scenarios based on the reference configuration results, and generates resource configuration results. The fault prevention submodule utilizes the resource configuration results to optimize power grid operating parameters and upgrade early warning systems, thereby obtaining fault prevention measures.

6. The power grid fault early warning system based on machine learning according to claim 5, characterized in that, The fault simulation and testing module includes a parameter tuning submodule, a strategy iteration submodule, and a strategy optimization and evaluation submodule, wherein... The parameter adjustment submodule performs virtual fault simulation based on the fault prevention measures to obtain simulation feedback optimization parameters; The parameter adjustment submodule constructs an operation parameter adjustment table based on the simulation feedback optimization parameters; The strategy iteration submodule uses the operation parameter adjustment table to optimize the fault response strategy and obtain the strategy optimization result. The strategy optimization and evaluation submodule uses the strategy optimization results to adjust the test parameters using a Bayesian optimization algorithm to obtain the test evaluation results.

7. The power grid fault early warning system based on machine learning according to claim 6, characterized in that, in, The strategy optimization and evaluation submodule implements the steps of adjusting test parameters using a Bayesian optimization algorithm based on the strategy optimization results to obtain test evaluation results through the following formula: μ(x)=k(x,θ) T (K(θ)+σ n 2 I+λL) -1 y Where μ(x) represents the predicted mean at position x, i.e., the test evaluation result; k(x,θ) represents the covariance vector between x adjusted according to parameter θ and the training data points; k(x,θ) T This represents the transpose of k(x,θ); K(θ) represents the covariance matrix between training data points adjusted according to the parameter θ; σ n λ represents the noise term; I represents the identity matrix; λ represents the regularization coefficient; L represents the Laplacian matrix; and y represents the observed values ​​of the objective function at the training data points.

8. The power grid fault early warning system based on machine learning according to any one of claims 1-7, characterized in that, The fault early warning and optimization module includes a fault detection submodule, an education update submodule, and a strategy update submodule, wherein... The fault detection submodule analyzes the operation logs based on the test evaluation results to obtain the abnormal frequency and abnormal operations. The fault detection submodule classifies and marks the anomalies based on the anomaly frequency and the anomaly operation, and obtains an anomaly classification and marking table; The education update submodule uses the anomaly classification label table to conduct training on fault response, and obtains an updated training material set. The strategy update submodule adjusts and optimizes the fault warning and response based on the updated training material set to obtain a reference fault warning and reference response strategy. The strategy update submodule performs simulation exercises and verifies the implementation effect of the reference response strategy based on the reference fault warning and the reference response strategy, and obtains the target fault warning and the target response strategy.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the processing method of the power grid fault early warning system based on machine learning as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the processing method of the power grid fault early warning system based on machine learning as described in any one of claims 1 to 8.

Images