Multi-view-angle-based power recovery method and system for power failure

Abstract

The invention discloses a multi-view-angle-based power recovery method and system for a power failure, and particularly relates to the technical field of power recovery of a power grid. The method comprises the following steps: acquiring multi-source data of a power system, acquiring multi-view time sequence data, performing feature extraction and classification by utilizing a power grid operation model based on time sequence analysis and combining a block type distributed load time sequence portrait method, and generating a node-level load dynamic evaluation result; based on an evaluation result, adopting a parallel computing strategy based on iterative optimization to identify a local overload risk, and generating a local priority list; based on the priority list, a node group power recovery sequence is set in a distributed scheduling mode, and a power supply line is recovered segment by segment through a parallel judgment circuit breaker control algorithm; in the power restoration process, based on a real-time monitoring result of a node load state, dynamically correcting a parallel computing strategy, and isolating and shunting load overrun and implicit sudden increase nodes; and the accuracy of the power recovery strategy and the safety of power grid operation are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid restoration, and more specifically, to a method and system for restoring power to power failures based on multiple perspectives. Background Art

[0002] In the prior art, during the power restoration process after a power failure, the dynamic behavior of the node load is usually simplified. However, some node loads may experience a "load shadow phenomenon" after power restoration, that is, there is a significant difference between the instantaneous behavior of the node load recovery and its actual stable state, which manifests as a hidden surge or delayed amplification. This phenomenon is difficult to capture through conventional monitoring methods and is easily overlooked in load dynamic assessment and overload risk identification.

[0003] When it is difficult to timely perceive and predict the dynamic characteristics of node loads, the power restoration strategy is likely to lead to local overload, power supply restoration delays and secondary failures during the power restoration process, which in turn has a serious impact on the safety and reliability of power grid operation. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a method and system for restoring power after a power failure based on multiple perspectives to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] The method for restoring power to a power failure based on multiple perspectives includes the following steps:

[0007] S1: Acquire multi-perspective time series data by collecting multi-source data of the power system;

[0008] S2: Use the power grid operation model based on time series analysis combined with the block-based distributed load time series profiling method to extract and classify the multi-view time series data to generate node-level load dynamic evaluation results;

[0009] S3: Based on the dynamic evaluation results of node-level load, a parallel computing strategy based on iterative optimization is used to identify possible local overload risks and output a local priority list;

[0010] S4: Based on the local priority list, a distributed scheduling method is used to set the power restoration order of the node group, and each power supply line is restored section by section through a circuit breaker control algorithm based on parallel judgment;

[0011] S5: Based on the real-time monitoring results of the load status of each node during the power restoration process, the iterative optimization parallel computing strategy is dynamically corrected, and the nodes with overload and hidden surges in the local priority list are re-isolated and diverted to complete orderly power restoration.

[0012] In a preferred embodiment, acquiring multi-view time series data by collecting multi-source data of the power system specifically includes:

[0013] The power grid monitoring system is used to obtain real-time data related to the fault location, the geographic information system is used to obtain spatial data related to the line topology, the synchronous phasor measurement unit is used to obtain node voltage, current and phasor timing data, and the edge computing node is used to obtain the real-time operation data of the load status of each node; the real-time data related to the fault location, the spatial data related to the line topology, the node voltage, current and phasor timing data, and the real-time operation data of the load status of each node are uniformly time-aligned and format-normalized to form multi-perspective timing data.

[0014] In a preferred embodiment, a power grid operation model based on time series analysis is used in combination with a block-based distributed load time series profiling method to extract and classify features of multi-view time series data to generate node-level load dynamic evaluation results, specifically including:

[0015] Through the power grid operation model of time series analysis, dynamic behavior modeling of multi-view time series data is carried out to extract the time series characteristics of node loads;

[0016] Using the block-based distributed load time series profiling method, the multi-view time series data is divided into multiple blocks, and pattern recognition and feature classification are performed on the data of each block to generate a time series profile containing the dynamic change characteristics of the node load;

[0017] The node load timing characteristics and the timing portrait containing the dynamic change characteristics of the node load are classified and aggregated to form the node-level load dynamic evaluation results.

[0018] In a preferred embodiment, the node load time series characteristics and the time series portrait containing the dynamic change characteristics of the node load are classified and aggregated to form a node-level load dynamic evaluation result, including:

[0019] The load characteristics and time series profile data of each node are classified, and the nodes with similar characteristics are classified into one category using a clustering algorithm. The clustering goal is to minimize the intra-class distance: Among them, min means that the goal is to minimize a certain value, ||X j -μ k || 2 Represents the jth block X j and the kth cluster center μ k The square of the Euclidean distance, X j Represents the feature vector contained in the time series image of the jth block, μ k represents the center vector of the kth cluster, K represents the total number of cluster categories, k represents the cluster number, C krepresents the set of blocks of the kth cluster;

[0020] All time series blocks belonging to the same node are aggregated, and according to the distribution characteristics of the blocks in different time periods, the node-level load dynamic evaluation results of the node are extracted, including the load fluctuation range, load change trend and load abnormal dynamic indicators.

[0021] In a preferred embodiment, based on the node-level load dynamic evaluation results, a parallel computing strategy based on iterative optimization is used to identify possible local overload risks and output a local priority list, specifically including:

[0022] Based on the node-level load dynamic assessment results, a load dynamic risk model is constructed to define the load overload risk factor for each node;

[0023] Initialize the load status data and related calculation parameters of each node, the related calculation parameters include the load reference value and the load range that can be carried;

[0024] Adopt parallel computing strategy for all nodes, iteratively optimize the load distribution between nodes, identify the nodes with local overload and their associations;

[0025] According to the node's load overlimit risk factor and the node-level load dynamic assessment results, a local priority list of nodes is generated.

[0026] In a preferred embodiment, based on the local priority list, a distributed scheduling method is used to set the power restoration order of the node group, and each power supply line is restored section by section through a circuit breaker control algorithm based on parallel judgment, specifically including:

[0027] Decompose the power restoration task into multiple node groups according to the local priority list, each node group includes several nodes and their corresponding power supply lines;

[0028] Based on the priority ranking within the node group, the power restoration order is set in sequence starting from the node group with the highest priority;

[0029] The circuit breaker status of each power supply line is judged in parallel to identify nodes where power supply has been restored and nodes where power supply has not been restored;

[0030] The power supply lines are gradually restored in the set power restoration order to ensure that the power supply lines of the node groups are connected to the main power grid in sequence.

[0031] In a preferred embodiment, based on the real-time monitoring results of the load status of each node during the power restoration process, the iterative optimization parallel computing strategy is dynamically corrected, and the nodes with overload and hidden surge in the local priority list are isolated and diverted again to complete orderly power restoration, which specifically includes:

[0032] Use edge computing nodes and online monitoring equipment to collect load data of each node in real time during the power restoration process, and monitor the load change trend, fluctuation amplitude and abnormal dynamic characteristics of the nodes;

[0033] Compare and analyze the node load status monitored in real time with the dynamic evaluation results of the nodes in the local priority list to identify the nodes with overload and hidden sudden increase in load in the local priority list;

[0034] Based on the analysis results, the load distribution rules between nodes in the iterative optimization parallel computing strategy are adjusted, and the load distribution values ​​of nodes with overload and hidden sudden increase are recalculated;

[0035] For nodes identified as abnormally loaded after recalculation, they are isolated and diverted by cutting off associated lines and adjusting distribution paths;

[0036] According to the revised calculation strategy and load distribution results, the local priority list is dynamically updated, and the power supply status of each node is restored in order of priority.

[0037] On the other hand, the present invention provides a power restoration system based on multi-perspective power failure, including a multi-source data acquisition module, a node feature classification module, an overload risk identification module, a power restoration sequence scheduling module, and a dynamic adjustment isolation module;

[0038] Multi-source data acquisition module: acquires multi-perspective time series data by collecting multi-source data of the power system;

[0039] Node feature classification module: It uses the power grid operation model based on time series analysis and the block-based distributed load time series profiling method to extract and classify the features of multi-view time series data and generate node-level load dynamic evaluation results;

[0040] Overload risk identification module: Based on the dynamic evaluation results of node-level load, it adopts a parallel computing strategy based on iterative optimization to identify possible local overload risks and output a local priority list;

[0041] Power restoration sequence scheduling module: Based on the local priority list, the power restoration sequence of the node group is set in a distributed scheduling manner, and each power supply line is restored section by section through a circuit breaker control algorithm based on parallel judgment;

[0042] Dynamically adjust the isolation module: Based on the real-time monitoring results of the load status of each node during the power restoration process, the iterative optimization parallel computing strategy is dynamically corrected, and the nodes with overload and hidden surges in the local priority list are re-isolated and diverted to complete orderly power restoration.

[0043] The technical effects and advantages of the power restoration method and system of the present invention based on multi-perspective power failure are as follows:

[0044] 1. By collecting multi-source data of the power system, combining time series analysis with block-based distributed load time series profiling methods, an accurate assessment of the dynamic behavior of node loads is achieved. Compared with the prior art, the present invention can comprehensively capture the dynamic characteristics of node loads, including the phenomena of implicit surges and delayed amplification, solving the problem that conventional monitoring methods are difficult to capture load shadowing phenomena. On this basis, through a parallel computing strategy based on iterative optimization, possible local overload risks are identified, and a local priority list is generated, providing more forward-looking and scientific guidance for scheduling and load management during the power restoration process.

[0045] 2. By further introducing real-time monitoring and dynamic adjustment mechanisms, based on the real-time monitoring results of the load status of each node during the power restoration process, the parallel computing strategy is dynamically corrected and iteratively optimized, and the nodes with overload or hidden sudden increase are isolated and diverted to ensure the load balance and node operation stability during the power restoration process. The power restoration order of the node group is set through a distributed scheduling method, and combined with the parallel judgment circuit breaker control algorithm, the stage-by-stage restoration of each power supply line is realized, effectively avoiding the occurrence of secondary faults. The present invention improves the accuracy of the power restoration strategy, the power restoration efficiency and the safety of power grid operation as a whole, and is particularly suitable for the refined management needs of power restoration scheduling in complex power grids. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic diagram of a method for restoring power to a power failure based on multiple perspectives of the present invention;

[0047] Figure 2 It is a structural schematic diagram of the power restoration system of the present invention based on multi-view power failure. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] Embodiment 1: Figure 1 The present invention provides a method for restoring power to a power failure based on multiple perspectives, which includes the following steps:

[0050] S1: Acquire multi-perspective time series data by collecting multi-source data of the power system.

[0051] S2: The power grid operation model based on time series analysis is combined with the block-based distributed load time series profiling method to extract and classify features of multi-perspective time series data to generate node-level load dynamic evaluation results.

[0052] S3: Based on the dynamic evaluation results of node-level load, a parallel computing strategy based on iterative optimization is used to identify possible local overload risks and output a local priority list.

[0053] S4: Based on the local priority list, a distributed scheduling method is used to set the power restoration order of the node group, and each power supply line is restored section by section through a circuit breaker control algorithm based on parallel judgment.

[0054] S5: Based on the real-time monitoring results of the load status of each node during the power restoration process, the iterative optimization parallel computing strategy is dynamically corrected, and the nodes with overload and hidden surges in the local priority list are re-isolated and diverted to complete orderly power restoration.

[0055] By collecting multi-source data of the power system, multi-perspective time series data can be obtained, including:

[0056] Use the power grid monitoring system to obtain real-time data related to the fault location:

[0057] The power grid monitoring system is used to obtain data related to the fault location in real time, including voltage drop at the fault point, equipment tripping signals and line protection action status.

[0058] First, the equipment operation status information and fault alarm information of each area in the power grid monitoring system are received; then, the data is collected at preset time intervals, and real-time data including the time when the alarm is triggered, the specific line equipment number and its fault type are recorded. This data is used to preliminarily determine the specific area where the fault occurs.

[0059] Use geographic information system to obtain spatial data related to line topology:

[0060] GIS provides spatial data related to the topology of the power grid, including line distribution, node locations and their spatial relationships with surrounding areas.

[0061] First, the spatial distribution map of lines and equipment is extracted by calling the geographic information interface of the power system; then, the power line topology of the current area is generated, including the associated information of trunk lines, branch lines and adjacent equipment. The collected spatial data provides support for subsequent fault location and line status analysis.

[0062] Use the synchronized phasor measurement unit to obtain node voltage, current and phasor timing data:

[0063] The synchronized phasor measurement unit is used to obtain the timing data of the voltage, current and phasor of the node, accurately reflecting the changes in the electrical characteristics between the nodes.

[0064] First, the voltage amplitude, current amplitude and phase angle of the node in the specified fault area are collected using the phasor measurement device; then, these time series data are recorded according to the timestamp to ensure that the sampling frequency meets the microsecond level accuracy. These data provide key support for the analysis of node dynamic characteristics and fault propagation paths.

[0065] Use edge computing nodes to obtain real-time operating data of the load status of each node:

[0066] Edge computing nodes are used to obtain real-time operating data of node load status, including dynamic changes in load power and changing trends in operating load.

[0067] First, the edge computing node collects real-time data such as load power and current fluctuations from various distributed sensors; then, the changing trend of the load status is converted into short-term forecast data to ensure the dynamic capture of load anomalies. These data provide the basis for subsequent load evaluation and priority list generation.

[0068] The real-time data related to the fault location, the spatial data related to the line topology, the node voltage, current and phasor time series data, and the real-time operation data of the load status of each node are uniformly time-aligned and format-normalized to form multi-view time series data:

[0069] First, the collected data are aligned according to the timestamps of each data source to ensure that the data from the power grid monitoring system, geographic information system, synchronized phasor measurement unit and edge computing node are in the same time base. Then, the data in different formats are standardized and converted, including the unification of numerical units, time resolution and data encoding format.

[0070] Finally, the processed data is constructed into a multi-dimensional time series data set, forming multi-perspective time series data including fault location, line topology, node electrical characteristics, and load status.

[0071] The power grid operation model based on time series analysis is combined with the block-based distributed load time series profiling method to extract and classify the multi-view time series data features and generate node-level load dynamic evaluation results, including:

[0072] The power grid operation model of time series analysis is used to model the dynamic behavior of multi-view time series data and extract the node load time series characteristics:

[0073] The collected multi-view time series data is denoised, and the high-frequency noise in the measurement data is eliminated by filtering. After denoising, the data is normalized so that the data from different sources have the same numerical range.

[0074] A time series prediction model is established for the load time series data of each node. Assume that the input multi-view time series data is {X1(t), X2(t), ..., X n (t)}; where X i (t) represents the load state of the i-th node at time t. The dynamic changes of node load are fitted by the time series model, and the expression is as follows: i (t) = α·X i (t-1)+β·X i (t-2)+γ; where Y i (t) represents the load forecast value of the i-th node at time t, which is used to reflect the dynamic load status of the node at the current moment; X i (t-1) represents the actual load value of the i-th node at time t-1, which is a record of the load status of the node at the previous moment; X i (t-2) represents the actual load value of the i-th node at time t-2, which is a record of the load status of the node at the previous two moments; α represents the X i (t-1) for Y i (t) is usually calculated by data fitting; β represents the X of the i-th node. i (t-2) for Y i (t), which is usually calculated by data fitting; γ represents the load baseline value of the i-th node, which is used to reflect the normal level of the node load and is usually a constant; t represents the current time point, which is used to distinguish the time dimension of the node load value; i represents the node number in the power grid, which is used to identify the specific node; n represents the total number of nodes in the power grid, which is used to define the scale of all nodes involved in the power grid load forecast.

[0075] The above time series model is used to calculate the periodicity and trend characteristics of each node, including the cycle length and trend direction of load changes. Statistical methods are used to identify abnormal fluctuation points, such as the moment when the load suddenly increases or decreases sharply.

[0076] Using the block-based distributed load time series profiling method, the multi-view time series data is divided into multiple blocks, and pattern recognition and feature classification are performed on the data of each block to generate a time series profile containing the dynamic change characteristics of the node load:

[0077] According to the time series characteristics of the nodes, the multi-perspective time series data is divided into multiple blocks. The length of the block is determined by the periodicity of the data changes. Assuming that the total duration of the time series data is T and the load cycle of the node is P, the length of each block is P, and a total of T / P blocks can be divided.

[0078] In each block, the dynamic change characteristics of the nodes are extracted, including the following three main features: the maximum, minimum and mean values ​​of the load within the block; the fluctuation amplitude of the load within the block; and the rate of change of the load within the block.

[0079] The rate of load change within a block is calculated by summing up the differences in load values ​​between adjacent time points and taking the average. The specific process is to compare the load value changes of all adjacent time points in the block one by one, accumulate these changes and divide them by the total number of time points in the block to obtain the average rate of load change in the block, which is used to measure the speed of load change.

[0080] The fluctuation amplitude of the load within the block is determined by calculating the difference between the maximum and minimum load values ​​within the block. The specific process is to traverse all load data points in the block, record the maximum load value and the minimum load value, and take the difference between the two as the amplitude of the load fluctuation within the block, which is used to reflect the size of the load change range.

[0081] The features of each block are used to generate a time series image according to the preset classification standard. The feature vector contained in each time series image is represented as {F j , R j , A j}, where F j represents the fluctuation amplitude of the load of the jth block, R j represents the load change rate of the jth block, A j It represents the load mean of the jth block, and j represents the block number.

[0082] The node load time series characteristics and the time series portraits containing the dynamic change characteristics of the node load are classified and aggregated to form the node-level load dynamic evaluation results:

[0083] The load characteristics and time series profile data of each node are classified, and the nodes with similar characteristics are classified into one category using a clustering algorithm. The clustering goal is to minimize the intra-class distance: Among them, min means that the goal is to minimize a certain value, which in this formula means minimizing the sum of the intra-class distances of all clusters; ||X j -μ k || 2 Represents the jth block X j and the kth cluster center μ k The square of the Euclidean distance is used to measure the similarity between the block and the cluster center; X j The feature vector contained in the time series portrait of the jth block contains the feature data of the block (such as fluctuation amplitude, change rate, mean value, etc.); μ krepresents the center vector of the kth cluster, which represents the average eigenvalue of the cluster and is used to represent the overall characteristics of the cluster; K represents the total number of cluster categories and is used to define the number of cluster divisions; k represents the cluster number; C k Represents the block set of the kth cluster, including all block numbers belonging to the cluster.

[0084] After clustering is completed, each C k Contains a group of time series block numbers j with similar dynamic characteristics. The feature vectors corresponding to these blocks are {F j ,R j ,A j}Describes the load characteristics of each block.

[0085] Each cluster center μ k As a representative feature of this class, the common characteristics of the blocks in this class are summarized, such as fluctuation amplitude, change rate and load mean.

[0086] Aggregate all time series blocks belonging to the same node, and extract the node-level load dynamic evaluation results of the node according to the distribution characteristics of the blocks in different time periods, including:

[0087] Load fluctuation range: the difference between the maximum and minimum load values ​​of the node load in all blocks;

[0088] Load change trend: the growth or decline trend of node load across multiple blocks;

[0089] Abnormal load dynamic indicators: Identify hidden sudden increases or delayed amplification behaviors of load based on clustering characteristics.

[0090] Based on the extracted node-level load dynamic evaluation results, a load dynamic evaluation report is generated for each node. The report content includes:

[0091] Node-level characteristics: the overall dynamic performance of the node, such as whether it is a high-volatility node or a load-stable node;

[0092] Time segment evaluation: Load characteristics in different time intervals, to identify time periods that may have load surges or abnormal behavior;

[0093] Risk classification: Based on dynamic characteristics and abnormal indicators, nodes are divided into normal nodes, high-risk nodes and nodes that need to be monitored.

[0094] Based on the dynamic evaluation results of node-level load, a parallel computing strategy based on iterative optimization is used to identify possible local overload risks and output a local priority list, including:

[0095] Based on the node-level load dynamic assessment results, a load dynamic risk model is constructed to define the load overload risk factor for each node:

[0096] Extract node-level load dynamic assessment results, including load fluctuation range, load change trend, and load abnormal dynamic indicators.

[0097] Define the load overload risk factor for each node as follows: Among them, R f Indicates the load overload risk factor of the node. The larger the value, the higher the overload risk. max Indicates the maximum load value of the node; L base Indicates the load baseline value of the node; C max Indicates the maximum load value that a node can bear.

[0098] The load overload risk factor of each node is calculated by the above formula and used for risk assessment and priority list generation in subsequent steps.

[0099] The risk factors are mapped with the node-level load dynamic assessment results, including load fluctuation range, load change trend and load abnormal dynamic indicators, to form a load dynamic risk model that can reflect the node risk status and provide basic data support for subsequent analysis.

[0100] Initialize the load status data and related calculation parameters of each node, including the load reference value and the load range that can be carried:

[0101] Load status data: Get the current actual load value of each node and record it in the load status table.

[0102] Define the load base value and the load range that can be carried:

[0103] Load baseline value: calculated based on the node's historical load data, serving as the node's normal operating load level;

[0104] Carryable load range: Calculated based on the design capacity and operating margin of the node, indicating the maximum load that the node can carry without any risk.

[0105] Construct a node status matrix to store the load status data and related calculation parameters of all nodes, and provide a unified data interface for subsequent calculations.

[0106] A parallel computing strategy is adopted for all nodes to iteratively optimize the load distribution between nodes and identify the nodes with local overload and their associations:

[0107] The load status data of each node is calculated in parallel, and the calculation formula is as follows: Among them, L dRepresents the load distribution value of the node, which is used to calculate the load level of the node under the current distribution strategy; L c Indicates the current load value of the node, reflecting the actual load of the current node; U represents the total number of nodes directly connected to the target node, which is used to quantify the load impact of neighboring nodes on the target node; L u Represents the load value of the uth directly connected node, which is used to evaluate the impact of the neighbor node load on the target node.

[0108] After each round of calculation, the node state matrix is ​​updated according to the load distribution result of the current node, and the load overload risk factor of each node is recalculated. The calculation is repeated until one of the following convergence conditions is met:

[0109] The load overload risk factor of all nodes is lower than the set threshold; the maximum number of iterations is reached.

[0110] In each iteration, the load change relationship between nodes is recorded and a load association graph is constructed to identify nodes with local load exceeding the limit and their association with other nodes.

[0111] Based on the node's load overload risk factor and the node-level load dynamic assessment results, a local priority list of nodes is generated:

[0112] Each node is sorted according to the load overload risk factor. The node with a higher load overload risk factor has a higher priority.

[0113] Combined with the dynamic characteristic data of the nodes (such as fluctuation range and change trend), the priority order is further adjusted to ensure that key nodes are processed first;

[0114] Output a priority list, including the following information: node number, node load overload risk factor, node dynamic characteristics description (such as fluctuation range, change trend, etc.) and node priority order.

[0115] Based on the local priority list, a distributed scheduling method is used to set the power restoration order of the node group, and the circuit breaker control algorithm based on parallel judgment is used to restore each power supply line section by section, including:

[0116] The power restoration task is decomposed into multiple node groups according to the local priority list. Each node group contains several nodes and their corresponding power supply lines:

[0117] Read the local priority list, which contains the priority, load status and corresponding power supply line information of each node; divide the power restoration task into several node groups according to the priority list, each node group consists of several nodes and their associated power supply lines; during the division process, ensure that the nodes in the same node group have adjacent topological structures so that power distribution can be effectively controlled during the recovery process; assign a unique number to each node group, and record the node numbers and power supply line information it contains to form a power restoration task table to provide input for subsequent steps.

[0118] Based on the priority order within the node group, the power restoration order is set starting from the node group with the highest priority:

[0119] The nodes in the node group are sorted according to the priority provided by the local priority list, and the nodes with higher priority are processed first. During the sorting process, if multiple nodes have the same priority, they are sorted in a secondary order according to the load characteristics of the nodes and the importance of the connection with the main power grid. According to the sorting results, the node group with the highest priority is set as the first power restoration target, and the power restoration order of other node groups is set in turn. The set power restoration order is recorded to form a power restoration plan table, which contains the node group number and its power restoration sequence number, providing a basis for subsequent circuit breaker control.

[0120] The circuit breaker status of each power supply line is judged in parallel to identify nodes where power supply has been restored and nodes where power supply has not been restored:

[0121] Read the power supply line information corresponding to each node group in the power restoration plan table to determine the list of circuit breakers that need to be checked; determine the status of each circuit breaker in the circuit breaker list to check whether the circuit breaker is in the "closed" or "open" state; determine whether the power supply line corresponding to the circuit breaker has been restored. If it has been restored, mark its associated node as "restored"; if not, mark it as "to be restored"; record the status of each circuit breaker and the recovery status of its associated nodes, generate a circuit breaker status table, and provide data support for subsequent power supply line restoration.

[0122] Restore each power supply line step by step according to the set power restoration sequence to ensure that the power supply lines of the node group are connected to the main power grid in sequence:

[0123] Starting from the node group with the highest priority, restore the power supply lines one by one in the order set in the power restoration plan; before restoring each power supply line, check whether the status of its associated circuit breaker meets the restoration conditions and ensure that the circuit breaker is in the "closed" state; send restoration instructions to the power supply lines that meet the conditions, restore the power supply of the lines section by section, and update the circuit breaker status table in real time after restoration; during the restoration process, record the power supply restoration status of each node group to ensure that all nodes and their power supply lines are connected to the main power grid.

[0124] Based on the real-time monitoring results of the load status of each node during the power restoration process, the parallel computing strategy is dynamically corrected and iteratively optimized, and the nodes with overload and hidden surge in the local priority list are isolated and diverted again to complete orderly power restoration. Specifically, it includes:

[0125] Use edge computing nodes and online monitoring equipment to collect real-time load data of each node during the power restoration process, and monitor the load change trend, fluctuation amplitude and abnormal dynamic characteristics of the nodes:

[0126] During the power restoration process, edge computing nodes and online monitoring equipment are used to collect load data of each node in real time so as to dynamically obtain the operating status of the node.

[0127] Each edge computing node is responsible for monitoring the node load within its control range, including the current load value, load change trend and historical load data; the online monitoring equipment collects the node's load fluctuation amplitude and abnormal load behavior (such as hidden surges or delayed amplification) in real time, and records the data to the real-time monitoring table; all monitoring data are uniformly formatted according to timestamps to ensure the time consistency and comparability of the data, providing accurate input for subsequent analysis.

[0128] Compare and analyze the real-time monitored node load status with the dynamic evaluation results of the nodes in the local priority list to identify the nodes with overload and hidden sudden increase in load in the local priority list:

[0129] According to the real-time monitoring table, extract the current load data of each node and compare it with the dynamic evaluation results of the nodes in the priority list item by item; focus on analyzing the differences between the current load data and the dynamic evaluation results, including the fluctuation range, change trend and deviation of abnormal dynamic indicators; mark the load-exceeding nodes and hidden surge nodes according to the analysis results, and add the identified abnormal nodes to the abnormal node list to provide a basis for subsequent correction strategies.

[0130] Based on the analysis results, the load distribution rules between nodes in the iterative optimization parallel computing strategy are adjusted, and the load distribution values ​​for nodes with overload and hidden sudden increase are recalculated:

[0131] The load status of abnormal nodes is remodeled, taking into account their current load level, the load status of neighboring nodes and the line capacity. The load distribution rules are adjusted according to the revised model to ensure that the load distribution is balanced among neighboring nodes. The load distribution values ​​of all nodes are recalculated using a parallel computing strategy, and the node status table is updated in real time.

[0132] For nodes that are identified as abnormally loaded after recalculation, they are isolated and diverted by cutting off associated lines and adjusting distribution paths:

[0133] For nodes marked as overloaded, they are temporarily isolated by cutting off their associated power supply lines; for nodes marked as hidden surges, their loads are diverted to neighboring nodes by adjusting the distribution path; after the processing is completed, the load status of the relevant nodes and the distribution relationship of their neighboring nodes in the node status table are updated to ensure data consistency after isolation and diversion operations.

[0134] According to the revised calculation strategy and load distribution results, the local priority list is dynamically updated, and the power supply status of each node is restored in order of priority:

[0135] Write the corrected load distribution value and node status into the local priority list and re-sort them according to the updated data; gradually restore the power supply status of the nodes in order of priority, starting from the node with the highest priority one by one; record the power supply status during the recovery process and the information of the remaining nodes to be restored, and ensure that the list content is updated in real time after the power is restored.

[0136] Embodiment 2: Embodiment 2 of the present invention differs from Embodiment 1 in that this embodiment introduces a power restoration system for power failures based on multiple perspectives.

[0137] Figure 2 A structural schematic diagram of the power restoration system based on multi-perspective power failure of the present invention is given. The power restoration system based on multi-perspective power failure includes a multi-source data acquisition module, a node feature classification module, an overload risk identification module, a power restoration sequence scheduling module and a dynamic adjustment isolation module.

[0138] Multi-source data acquisition module: acquires multi-perspective time series data by collecting multi-source data of the power system.

[0139] Node feature classification module: The power grid operation model based on time series analysis is combined with the block-based distributed load time series portrait method to extract and classify features of multi-perspective time series data to generate node-level load dynamic evaluation results.

[0140] Overload risk identification module: Based on the dynamic evaluation results of node-level load, a parallel computing strategy based on iterative optimization is used to identify possible local overload risks and output a local priority list.

[0141] Power restoration sequence scheduling module: Based on the local priority list, a distributed scheduling method is used to set the power restoration sequence of the node group, and each power supply line is restored section by section through a circuit breaker control algorithm based on parallel judgment.

[0142] Dynamically adjust the isolation module: Based on the real-time monitoring results of the load status of each node during the power restoration process, the iterative optimization parallel computing strategy is dynamically corrected, and the nodes with overload and hidden surges in the local priority list are re-isolated and diverted to complete orderly power restoration.

[0143] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0144] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0145] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0146] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0147] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0148] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0149] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0150] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0151] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0152] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for restoring power to power failure based on multiple perspectives, characterized in that: The steps include: S1: Acquire multi-perspective time series data by collecting multi-source data of the power system; S2: Use the power grid operation model based on time series analysis combined with the block-based distributed load time series profiling method to extract and classify the multi-view time series data to generate node-level load dynamic evaluation results; S3: Based on the dynamic evaluation results of node-level load, a parallel computing strategy based on iterative optimization is used to identify possible local overload risks and output a local priority list; S4: Based on the local priority list, a distributed scheduling method is used to set the power restoration order of the node group, and each power supply line is restored section by section through a circuit breaker control algorithm based on parallel judgment; S5: Based on the real-time monitoring results of the load status of each node during the power restoration process, the iterative optimization parallel computing strategy is dynamically corrected, and the nodes with overload and hidden surges in the local priority list are re-isolated and diverted to complete orderly power restoration.

2. The method for restoring power to a power failure based on multiple perspectives according to claim 1, characterized in that: By collecting multi-source data of the power system, multi-perspective time series data can be obtained, including: The power grid monitoring system is used to obtain real-time data related to the fault location, the geographic information system is used to obtain spatial data related to the line topology, the synchronous phasor measurement unit is used to obtain node voltage, current and phasor timing data, and the edge computing node is used to obtain the real-time operation data of the load status of each node; the real-time data related to the fault location, the spatial data related to the line topology, the node voltage, current and phasor timing data, and the real-time operation data of the load status of each node are uniformly time-aligned and format-normalized to form multi-perspective timing data.

3. The method for restoring power to a power failure based on multiple perspectives according to claim 1, characterized in that: The power grid operation model based on time series analysis is combined with the block-based distributed load time series profiling method to extract and classify the multi-view time series data features and generate node-level load dynamic evaluation results, including: Through the power grid operation model of time series analysis, dynamic behavior modeling of multi-view time series data is carried out to extract the time series characteristics of node loads; Using the block-based distributed load time series profiling method, the multi-view time series data is divided into multiple blocks, and pattern recognition and feature classification are performed on the data of each block to generate a time series profile containing the dynamic change characteristics of the node load; The node load timing characteristics and the timing portrait containing the dynamic change characteristics of the node load are classified and aggregated to form the node-level load dynamic evaluation results.

4. The method for restoring power to a power failure based on multiple perspectives according to claim 3, characterized in that: The node load time series characteristics and the time series portraits containing the dynamic change characteristics of the node load are classified and aggregated to form the node-level load dynamic evaluation results, including: The load characteristics and time series profile data of each node are classified, and the nodes with similar characteristics are classified into one category using a clustering algorithm. The clustering goal is to minimize the intra-class distance: Among them, min means that the goal is to minimize a certain value, ||X j -μ k || 2 Represents the jth block X j and the kth cluster center μ k The square of the Euclidean distance, X j Represents the feature vector contained in the time series image of the jth block, μ k represents the center vector of the kth cluster, K represents the total number of cluster categories, k represents the cluster number, C k represents the set of blocks of the kth cluster; All time series blocks belonging to the same node are aggregated, and according to the distribution characteristics of the blocks in different time periods, the node-level load dynamic evaluation results of the node are extracted, including the load fluctuation range, load change trend and load abnormal dynamic indicators.

5. The method for restoring power to a power failure based on multiple perspectives according to claim 1, characterized in that: Based on the dynamic evaluation results of node-level load, a parallel computing strategy based on iterative optimization is used to identify possible local overload risks and output a local priority list, including: Based on the node-level load dynamic assessment results, a load dynamic risk model is constructed to define the load overload risk factor for each node; Initialize the load status data and related calculation parameters of each node, the related calculation parameters include the load reference value and the load range that can be carried; Adopt parallel computing strategy for all nodes, iteratively optimize the load distribution between nodes, identify the nodes with local overload and their associations; According to the node's load overlimit risk factor and the node-level load dynamic assessment results, a local priority list of nodes is generated.

6. The method for restoring power to a power failure based on multiple perspectives according to claim 1, characterized in that: Based on the local priority list, a distributed scheduling method is used to set the power restoration order of the node group, and the circuit breaker control algorithm based on parallel judgment is used to restore each power supply line section by section, including: Decompose the power restoration task into multiple node groups according to the local priority list, each node group includes several nodes and their corresponding power supply lines; Based on the priority ranking within the node group, the power restoration order is set in sequence starting from the node group with the highest priority; The circuit breaker status of each power supply line is judged in parallel to identify nodes where power supply has been restored and nodes where power supply has not been restored; The power supply lines are gradually restored in the set power restoration order to ensure that the power supply lines of the node groups are connected to the main power grid in sequence.

7. The method for restoring power from a power failure based on multiple perspectives according to claim 1, characterized in that: Based on the real-time monitoring results of the load status of each node during the power restoration process, the parallel computing strategy is dynamically corrected and iteratively optimized, and the nodes with overload and hidden surge in the local priority list are isolated and diverted again to complete orderly power restoration. Specifically, it includes: Use edge computing nodes and online monitoring equipment to collect real-time load data of each node during the power restoration process, and monitor the load change trend, fluctuation amplitude and abnormal dynamic characteristics of the nodes; Compare and analyze the node load status monitored in real time with the dynamic evaluation results of the nodes in the local priority list to identify the nodes with overload and hidden sudden increase in load in the local priority list; Based on the analysis results, the load distribution rules between nodes in the iterative optimization parallel computing strategy are adjusted, and the load distribution values ​​of nodes with overload and hidden sudden increase are recalculated; For nodes identified as abnormally loaded after recalculation, they are isolated and diverted by cutting off associated lines and adjusting distribution paths; According to the revised calculation strategy and load distribution results, the local priority list is dynamically updated, and the power supply status of each node is restored in order of priority.

8. A power restoration system based on multi-perspective power failure, used to implement the power restoration method based on multi-perspective power failure according to any one of claims 1 to 7, characterized in that: It includes multi-source data acquisition module, node feature classification module, overload risk identification module, power restoration sequence scheduling module and dynamic adjustment isolation module; Multi-source data acquisition module: acquires multi-perspective time series data by collecting multi-source data of the power system; Node feature classification module: It uses the power grid operation model based on time series analysis and the block-based distributed load time series profiling method to extract and classify the features of multi-view time series data and generate node-level load dynamic evaluation results; Overload risk identification module: Based on the dynamic evaluation results of node-level load, it adopts a parallel computing strategy based on iterative optimization to identify possible local overload risks and output a local priority list; Power restoration sequence scheduling module: Based on the local priority list, the power restoration sequence of the node group is set in a distributed scheduling manner, and each power supply line is restored section by section through a circuit breaker control algorithm based on parallel judgment; Dynamically adjust the isolation module: Based on the real-time monitoring results of the load status of each node during the power restoration process, the iterative optimization parallel computing strategy is dynamically corrected, and the nodes with overload and hidden surges in the local priority list are re-isolated and diverted to complete orderly power restoration.

Images