Regional power grid fault processing decision generation method based on knowledge graph

Abstract

The invention provides a regional power grid fault processing decision generation method based on a knowledge graph, and relates to the technical field of power grid fault processing, and the method comprises the steps: constructing the knowledge graph covering multiple aspects of knowledge such as power grid equipment, operation states, fault cases and the like, and building the association relationship between the knowledge through integrating various data resources. When a fault occurs, a fault point is quickly and accurately positioned by utilizing the reasoning capability of the knowledge graph, and a scientific and reasonable fault processing decision scheme is generated in combination with historical fault processing experience and real-time operation data. According to the method, the fault processing efficiency and accuracy are improved, the operation safety and stability of the regional power grid are enhanced, new technical support and solution are provided for efficient operation and maintenance of the smart power grid, and the method has important theoretical significance and practical application value.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid fault processing, and in particular to a method for generating regional power grid fault processing decisions based on a knowledge graph. Background Art

[0002] With the construction and development of smart grids, the management and operation and maintenance models of power systems also need to be continuously innovated and upgraded. In actual use, the current knowledge graph platforms on the power grid on the market simply convert the power grid structure into a knowledge graph form; in short, it just changes to a more intuitive way of expression. In the actual fault handling process, these methods do not provide much support for the efficiency and accuracy of the fault handling solution, nor do they take into account the safety and stability of the power grid operation, as well as the management and operation and maintenance of the power system.

[0003] In view of this, this application is filed. Summary of the invention

[0004] The present invention provides a method for generating regional power grid fault handling decisions based on a knowledge graph, which can at least partially improve the above-mentioned problems.

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

[0006] A method for generating regional power grid fault processing decisions based on knowledge graph, comprising:

[0007] Acquire the current regional power grid structure to be processed, perform equivalent processing on the regional power grid structure using the knowledge graph data structure, and generate a knowledge graph equivalent structure of the regional power grid, wherein the knowledge graph equivalent structure includes nodes and edges;

[0008] Acquire monitoring data of the regional power grid structure collected by a preset collection module, and preprocess the monitoring data to obtain a preprocessed monitoring data sequence, wherein the preprocessing includes data filtering processing and normalization processing;

[0009] Calculating the Pearson correlation coefficient between the plurality of preprocessed monitoring data sequences, and preliminarily determining the fault area according to the Pearson correlation coefficient to obtain a preliminary fault area monitoring point set;

[0010] Based on the knowledge graph equivalent structure, the preliminary fault area monitoring point set is searched to determine the fault area node set, the fault suspicion of each node in the fault area node set is calculated, and the most likely fault area position is determined according to the magnitude of each fault suspicion;

[0011] Based on the preset fault pair key performance indicators of the power grid, the most likely fault area location is evaluated to determine the severity of the fault, and according to the edges of the knowledge graph equivalent structure, a breadth-first search process or a depth-first search process is performed on the set of nodes in the fault area to obtain a collection of affected nodes to determine the range of the most likely fault area location;

[0012] According to the fault severity of the most likely fault area location and the affected node set, a comprehensive evaluation function for the most likely fault area location is constructed, a plurality of processing solutions related to the most likely fault area location are extracted from a preset memory, and each processing solution is calculated using the comprehensive evaluation function to obtain a plurality of corresponding comprehensive evaluation function values;

[0013] Each of the comprehensive evaluation function values ​​is screened, and a processing solution with the smallest comprehensive evaluation function value is screened out as the optimal solution.

[0014] In summary, the method for generating regional power grid fault handling decisions based on knowledge graphs includes: constructing a knowledge graph covering multiple aspects of knowledge such as power grid equipment, operating status, and fault cases, and establishing associations between knowledge by integrating various data resources. When a fault occurs, the reasoning ability of the knowledge graph is used to quickly and accurately locate the fault point, and a scientific and reasonable fault handling decision plan is generated by combining historical fault handling experience and real-time operation data. It aims to improve the efficiency and accuracy of fault handling, enhance the safety and stability of regional power grid operation, and provide new technical support and solutions for the efficient operation and maintenance of smart grids. It has important theoretical significance and practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flow chart of a method for generating regional power grid fault processing decisions based on a knowledge graph provided by an embodiment of the present invention;

[0016] Figure 2 It is a schematic diagram of an equivalent structure of a knowledge graph of a regional power grid provided by an embodiment of the present invention;

[0017] Figure 3 It is a schematic diagram of marking fault nodes in a power grid based on a fault handling decision method provided by an embodiment of the present invention;

[0018] Figure 4 It is a schematic diagram of scoring a processing solution based on a fault handling decision method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0020] With the continuous expansion of regional power grids and the increasing requirements for power supply reliability, traditional fault handling methods face many challenges, and the management and operation and maintenance models of power systems also need to be continuously innovated and upgraded. Therefore, the present invention provides a regional power grid fault handling decision generation method based on knowledge graph to provide strong support for the realization of intelligent management and operation of power systems.

[0021] Specifically, refer to Figure 1 As shown, the first embodiment of the present invention discloses a method for generating a regional power grid fault processing decision based on a knowledge graph, which can be executed by a regional power grid fault processing decision generating device based on a knowledge graph (hereinafter referred to as a generating device), in particular, by one or more processors in the generating device to implement the following method:

[0022] S1, obtaining the regional power grid structure to be processed, performing equivalent processing on the regional power grid structure using a knowledge graph data structure, and generating a knowledge graph equivalent structure of the regional power grid, wherein the knowledge graph equivalent structure includes nodes and edges;

[0023] Preferably, the knowledge graph equivalent structure includes: equating the equipment in the regional power grid structure to nodes in the knowledge graph equivalent structure, and the set of nodes is N = {n1, n2, n3, ..., n m}, m is the total number of nodes;

[0024] Among them, each node has a corresponding attribute set A i ={a i1 , a i2 , a i3 , ..., a in}, A i is the attribute set of the i-th node, a i1 is the first attribute of the i-th node, a in is the nth attribute of the i-th node;

[0025] The relationship between devices in the regional power grid structure is equivalent to the edges in the knowledge graph equivalent structure. The set of edges is: E = {e1, e2, e3, ..., e k}, where k is the total number of edges;

[0026] Use triples to represent each edge and its two corresponding nodes. The expression is (n p , r j , nq ), where r j is the relationship type, n p is the pth node, n q is the qth node.

[0027] In this embodiment, first, the knowledge graph technology is used to equivalently represent the current regional power grid architecture model, that is, the regional power grid structure. Specifically, the knowledge graph is used to equivalently represent the regional power grid architecture model, which includes node representation and edge representation. Entities such as equipment and nodes in the power grid are nodes in the knowledge graph; the set is N = {n1, n2, n3, ..., n m}, m is the total number of nodes. The edge represents the relationship between entities, and the set of edges is: E = {e1, e2, e3, ..., e k}, where k is the total number of edges.

[0028] Each node has a corresponding set A of attributes i ={a i1 , a i2 , a i3 , ..., a in}, the collection content includes attributes such as device type, capacity, and location. Each edge r j Connect two nodes n p and n q , can be expressed as a triple (n p , r j , n q ), where r j Indicates the relationship type, such as connection relationship, control relationship, etc.

[0029] S2, acquiring the monitoring data of the regional power grid structure collected by the preset collection module, and preprocessing the monitoring data to obtain a preprocessed monitoring data sequence, wherein the preprocessing includes data filtering processing and normalization processing;

[0030] Specifically, step S2 includes: setting u monitoring points for the regional power grid structure, wherein the acquisition module can collect z different types of physical quantities at each monitoring point, and all the collected physical quantities constitute a monitoring data;

[0031] Acquire the monitoring data collected by the acquisition module, wherein the monitoring data is represented by a u×z matrix D, D ij is the value sequence of the jth physical quantity collected at the i-th monitoring point, i = 1, 2, ..., u, j = 1, 2, ..., z, D ij (T) is the sequence of the jth physical quantity collected by the i-th monitoring point in the observation time period T, D ij(t) is the value of the jth physical quantity collected by the i-th monitoring point at time t;

[0032] Based on digital filtering technology, a moving average filter is used to filter the sequence D of the jth physical quantity collected at the i-th monitoring point within time t. ij (T) is removed to obtain the filtered value Where B is the size of the moving average window;

[0033] The filtered value Normalize the filtered value Normalize to the preset range, where the normalization process uses the minimum-maximum normalization process, and the formula is: in, is the sequence of the jth physical quantity collected from the ith monitoring point after preprocessing, is the minimum value of the jth physical quantity collected at the ith monitoring point during the forecast period, is the maximum value of the jth physical quantity collected by the ith monitoring point within the preset time period.

[0034] In this embodiment, the voltage and current data of the power grid, i.e., the monitoring data, are collected in real time and preprocessed. For u monitoring points in the target power grid, z different types of physical quantities, such as current, voltage, power, etc., are collected for each monitoring point. In this embodiment, the monitoring data can be represented by a matrix D of u×z. In order to further consider the time characteristics of the data, the data sequence collected within a period of time is represented as D ij (T).

[0035] Subsequently, the collected monitoring data is preprocessed, including data filtering and normalization. First, digital filtering technology is used to remove noise interference, that is, a moving average filter is used to filter the monitoring data sequence D. ij (T) is filtered, and the filtered value It can be expressed by the following formula: Where B is the size of the moving average window.

[0036] After filtering, the data is normalized to a preset range. The minimum-maximum normalization method can be used, and its formula is expressed as

[0037] S3, calculating the Pearson correlation coefficient between the plurality of preprocessed monitoring data sequences, and preliminarily determining the fault area according to the Pearson correlation coefficient, to obtain a preliminary fault area monitoring point set;

[0038] Specifically, step S3 includes: obtaining the preprocessed monitoring data sequence and Calculate monitoring data series and Pearson correlation coefficient r iv , where v = 1, 2, ..., u, i ≠ v;

[0039] Pearson correlation coefficient r is The calculation formula is:

[0040] Determine the Pearson correlation coefficient r iv Whether it exceeds the preset threshold r th ;

[0041] If so, it is determined that there is a related fault between the i-th monitoring point and the v-th monitoring point, and the initial fault area monitoring point set M = {i1, i2, ..., i o}, i1 is the first preliminary fault area monitoring point, i o is the oth monitoring point in the preliminary fault area, and o is the number of monitoring points in the preliminary fault area;

[0042] If not, it is determined that there is no associated fault between the i-th monitoring point and the v-th monitoring point.

[0043] S4, searching the preliminary fault area monitoring point set based on the knowledge graph equivalent structure, determining the fault area node set, calculating the fault suspicion of each node in the fault area node set, and identifying the most likely fault area location according to the magnitude of each fault suspicion;

[0044] Specifically, step S4 includes: searching and reasoning in the knowledge graph equivalent structure for nodes corresponding to all monitoring points in the preliminary fault area monitoring point set to obtain the fault area node set N F ={n f1 , n f2 , ..., n fb},n f1 is the first fault region node, n fb is the bth fault area node;

[0045] According to the edges of the equivalent structure of the knowledge graph and the preset electrical principle knowledge, each node in the fault area node set is calculated to calculate its fault suspicion S(n fi ), the formula is: Among them, S(n fi ) is the fault suspicion of the node in the i-th fault area, (n fi ,r j ,n fj) is the i-th fault region node, the j-th fault region node and the relationship type r j’ The triplet of w j′ For the relationship type r j’ The importance of fault propagation and the probability of fault propagation determine the weight coefficient, I fj′ (t) is the fault judgment index corresponding to the monitoring point j′;

[0046] From the suspected faults of all fault area nodes, select one or more fault area nodes with the highest suspected faults, and identify the node as the most likely fault area location. The formula is: fault = arg max nfi∈NF S(n fi ), n fault The most likely fault area location.

[0047] In this embodiment, step S3 and step S4 can be understood as performing a correlation analysis on the preprocessed detection data to preliminarily determine the fault area; and then accurately locating the preliminarily determined fault area based on the knowledge graph equivalent structure.

[0048] The first step is to preliminarily determine the fault area by calculating the correlation coefficient of the data between different monitoring points; and Calculate the Pearson correlation coefficient r between them iv , the formula is: When the correlation coefficient exceeds the preset normal range, the threshold can be set based on experience, and it is considered that there may be a related fault between the two monitoring points. In this way, a region that may contain a fault can be preliminarily determined, and the set of monitoring points in the region is recorded as M = {i1, i2, ..., i o}.

[0049] In the second step, for the equipment corresponding to the monitoring points in the initially determined fault area, search and reason in the equivalent structure of the knowledge graph. The node set in the fault area can be represented as N in the knowledge graph. F ={n f1 , n f2 , ..., n fb Then, the fault suspicion of each node is calculated based on the connection relationship and electrical principle in the knowledge graph. fi , if it has direct connections with multiple nodes with abnormal monitoring data, and these connections have a high probability of causing fault propagation in the fault propagation model, then its fault suspicion is high, and the fault suspicion can be calculated by the following mathematical formula: Therefore, the node or node set with the highest suspected fault can be considered as the most likely fault location, and its mathematical expression is:

[0050] S5, based on the preset fault key performance indicators of the power grid, the most likely fault area location is evaluated to determine the severity of the fault, and according to the edges of the knowledge graph equivalent structure, a breadth-first search or a depth-first search is performed on the set of nodes in the fault area to obtain a collection of affected nodes to determine the range of the most likely fault area location;

[0051] Specifically, step S5 includes: evaluating the most likely fault area location based on preset fault key performance indicators of the power grid, evaluating the fault severity, and the fault severity function FS(n fi ) is: w h is the weight of the hth key performance indicator, and the key performance indicator vector for normal operation of the power grid is KPI normal ={kpi n1 ,kpi n2 ,...,kpi nr′}, the key performance indicator vector when the fault occurs is KPI fault ={kpi f1 ,kpi f2 ,...,kpi fr′}, r′ is the number of key performance indicator vectors;

[0052] According to the edges of the knowledge graph equivalent structure, a breadth-first search or a depth-first search is performed on each fault region node in the fault region node set to obtain the affected node set IR(n fi ), specifically:

[0053] Assume that the adjacency matrix of the knowledge graph equivalent structure is A1, the search depth threshold is d1, and the initialization queue Q = {n fi}, the visited node set Visited = {n fi};

[0054] When it is determined during the search process that the queue Q is not empty, the first node n of the queue Q and all nodes n' that are connected to the first node n by edges and have not been visited are extracted;

[0055] Add node n' to the known queue Q and the visited node set Visited;

[0056] Repeat the above steps until the search depth reaches the search depth threshold d1, and obtain the affected node set IR(n fi )=Visited.

[0057] In this embodiment, after confirming the fault area, it is necessary to further evaluate the severity and scope of the fault in the located area. The fault severity can be evaluated based on the impact of the fault on the key performance indicators of the power grid. The key performance indicators of the power grid mentioned here include: power transmission capability, voltage stability, etc. The fault severity function is defined as: Among them, w h is the weight of the hth key performance indicator, which can be determined according to its importance to the power grid operation.

[0058] Furthermore, in order to determine the impact range of the fault, it is necessary to use the connection relationship in the knowledge graph to identify the fault node n. fi Start a breadth-first search or depth-first search to calculate the set of affected nodes IR(n fi ), specifically:

[0059] Assume that the adjacency matrix of the knowledge graph is A1, and the search depth threshold is d1; taking breadth-first search as an example, initialize the queue Q = {n fi}、Visited node set Visited={n fi During the search process, when the queue Q is not empty, the first node n is taken out, and for all nodes n' that are connected to the first node n and have not been visited, the mathematical formula can be expressed as: A1[n][n']=1 and Add node n' to the queue Q and the visited node set Visited. Repeat this process until the search depth reaches d1, and finally get IR(n fi )=Visited, the affected node set can be determined.

[0060] S6, constructing a comprehensive evaluation function for the most likely fault area location according to the fault severity of the most likely fault area location and the affected node set, extracting multiple processing solutions related to the most likely fault area location from a preset memory, and using the comprehensive evaluation function to calculate each processing solution to obtain multiple corresponding comprehensive evaluation function values;

[0061] Specifically, step S6 includes: constructing a comprehensive evaluation function G(P i ), wherein the comprehensive evaluation function G(P i ) is the weighted sum of the recovery time, cost and grid stability impact of the treatment solution, and the formula is G(P i )=w1t recovery (P i )+w2c(P i )+w3s stability (P i), w1 is the weight of recovery time, w2 is the weight of cost, w3 is the weight of grid stability, w1+w2+w3=1, P i is the i-th processing solution stored in the preset memory;

[0062] Among them, for each processing solution, it is assumed that it contains X operation steps, and the estimated execution time of each operation step is t iJ (J = 1, 2, ..., X), and use the directed acyclic graph DAG to represent the order of the operation steps. Let its adjacency matrix be B1. If operation step J must be performed after step K is completed, B1 JK = 1, if operation step J does not have to be performed after step K is completed, B1 JK =0;

[0063] Calculate the earliest start time EST for each operation step J , EFT J =EST JJ +t iJ , among which, EFT J is the earliest completion time, EST1=0, EST1 is the first operation step;

[0064] When J>1, EFT K is the time of the Kth operation step;

[0065] Solution P i The recovery time is the earliest completion time of the last operation step, and we get t recovery (P i )=EFT X .

[0066] Solution P i The cost c(P i ) is composed of equipment maintenance cost C repair (P i ), Human resource cost C human (P i ) and power outage loss cost C power-loss (P i )composition;

[0067] Among them, suppose the set of equipment that needs to be repaired or replaced is L i = {l i1 , l i2 , ..., l iR}, the cost of repair or replacement of each device is c lj , get the equipment maintenance cost

[0068] Set the number of people involved in troubleshooting to npeople , the working hours of each person are H, and the unit time wage is W hour , get the human resource cost C human (P i )=n people ×H×W hour ;

[0069] Assume the total load in the blackout area is L load , the power outage time is t power-loss , the unit power outage loss is C loss-per-unit , and the power outage loss cost C is obtained power-loss (P i )=L load ×t power-loss ×C loss-per-unit ;

[0070] Get the treatment plan P i The cost c(P i )=C repair (P i )+C human (P i )+C power-loss (P i ).

[0071] The grid stability impact of the treatment solution is measured by the voltage stability index VSI and the frequency stability index FSI, specifically:

[0072] In the treatment plan P i After the execution is completed, the voltage stability index is obtained And frequency stability index And they are respectively the treatment scheme P i Voltage stability index And treatment plan P i Frequency stability index Assign weights w4 and w5, where w4+w5=1;

[0073] Get the treatment plan P i Impact on grid stability Among them, VSI normal FSI is the voltage stability index during normal operation. normal It is the frequency stability index.

[0074] In this embodiment, the calculation steps of the recovery time are as follows: for each processing scheme, assuming that it contains X operation steps, the estimated execution time of each operation step is t iJ(J = 1, 2, ..., X); At the same time, considering the order dependency between the operation steps, a directed acyclic graph DAG is used to represent the order of the operation steps. Let its adjacency matrix be B1. If operation step J must be performed after operation step K is completed, then B1 JK =1, otherwise B1 JK = 0. Calculate the earliest start time EST for each operation step J and earliest completion time EFT J , where it is assumed that the first operation step can start immediately, that is, EST1 = 0. Earliest completion time EFT J The formula is: EFT J =EST J +t iJ ; When J>1, it satisfies It can be seen that the treatment plan P i The recovery time is the earliest completion time of the last operation step, that is, t recovery (P i )=EFT X .

[0075] In this embodiment, the cost c(P i ) The calculation steps are: calculate the equipment maintenance cost, human resource cost and power outage loss cost in turn, and the sum of the three is the treatment plan P i The cost c(P i ). Impact on grid stability stability (P i ) The calculation steps are: When executing the processing scheme P i After that, the voltage stability index is obtained respectively. And frequency stability index According to the formula Calculate grid stability impact.

[0076] S7, screening each of the comprehensive evaluation function values, and screening out the processing scheme with the smallest comprehensive evaluation function value as the optimal scheme.

[0077] Preferably, the formula of the optimal solution is:

[0078] In this embodiment, the comprehensive evaluation function value of each processing scheme is calculated, and the processing scheme with the smallest comprehensive evaluation function value is selected as the optimal scheme.

[0079] Specifically, in this embodiment, in order to verify the effectiveness of the solution proposed in the present invention, a complex regional power grid architecture is constructed in MATLAB to verify how to make fault handling decisions based on real-time data. By defining the power grid structure, collecting voltage and current data, locating the fault node, evaluating the severity of the fault, and selecting a handling solution, the results are finally presented. Specifically, Figure 2 The overall structure of the power grid is shown, including 8 nodes (AH) and the connections (edges) between them. Each node represents an electric device or distribution point in the power grid. Figure 2 As you can see, node A is the master node, directly connected to nodes B and C, node B is connected to D and E, node C is connected to F, and so on for the other nodes.

[0080] See also Figure 3 , Figure 3 Based on the grid architecture model, the faulty nodes are marked (highlighted in red). In this case, node G is identified as a faulty node because the voltage drops below the threshold. With the proposed fault decision method, the problems in the grid can be quickly located, so that corresponding repair measures can be taken.

[0081] See also Figure 4 , Figure 4 The scoring results of three different treatment plans are shown. Each bar represents the score of a treatment plan. The higher the score, the more effective the plan is in restoring grid operation. By comparison, the best treatment plan can be quickly identified and a decision can be made. Among them, Plan 1 is a node isolation and load transfer plan, Plan 2 is a fault node repair plan, and Plan 3 is an increase in backup power supply / external power input plan.

[0082] In summary, the regional power grid fault handling decision generation method based on knowledge graph can organically integrate and manage various knowledge and information in the power system by constructing a knowledge graph to form a unified knowledge platform. This platform can not only be used for fault handling decisions, but also provide services for other business areas such as power grid planning and design, operation and dispatching, and equipment management. For example, in terms of equipment management, the knowledge graph can record the full life cycle information of the equipment, including data and knowledge of each link such as procurement, installation, commissioning, operation, maintenance, overhaul, and scrapping. Through the analysis and mining of this information, the intelligent operation and maintenance and status evaluation of the equipment can be realized, the reliability and service life of the equipment can be improved, and the operation and maintenance costs can be reduced. At the same time, the knowledge graph can also provide decision support for the operation and dispatch of the power system, optimize the operation mode of the power grid, improve energy utilization efficiency, and realize the intelligent and refined management of the power system. The regional power grid fault handling decision generation method based on knowledge graph can improve the efficiency and accuracy of fault handling, enhance the safety and stability of power grid operation, promote the intelligent management and operation and maintenance of the power system, and provide strong support for ensuring the reliable power supply of the regional power grid and the sustainable development of the social economy.

[0083] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for generating regional power grid fault handling decisions based on knowledge graph, characterized in that: include: Acquire the current regional power grid structure to be processed, perform equivalent processing on the regional power grid structure using the knowledge graph data structure, and generate a knowledge graph equivalent structure of the regional power grid, wherein the knowledge graph equivalent structure includes nodes and edges; Acquire monitoring data of the regional power grid structure collected by a preset collection module, and preprocess the monitoring data to obtain a preprocessed monitoring data sequence, wherein the preprocessing includes data filtering processing and normalization processing; Calculating the Pearson correlation coefficient between the plurality of preprocessed monitoring data sequences, and preliminarily determining the fault area according to the Pearson correlation coefficient to obtain a preliminary fault area monitoring point set; Based on the knowledge graph equivalent structure, the preliminary fault area monitoring point set is searched to determine the fault area node set, the fault suspicion of each node in the fault area node set is calculated, and the most likely fault area position is determined according to the magnitude of each fault suspicion; Based on the preset fault pair key performance indicators of the power grid, the most likely fault area location is evaluated to determine the severity of the fault, and according to the edges of the knowledge graph equivalent structure, a breadth-first search process or a depth-first search process is performed on the set of nodes in the fault area to obtain a collection of affected nodes to determine the range of the most likely fault area location; According to the fault severity of the most likely fault area location and the affected node set, a comprehensive evaluation function for the most likely fault area location is constructed, a plurality of processing solutions related to the most likely fault area location are extracted from a preset memory, and each processing solution is calculated using the comprehensive evaluation function to obtain a plurality of corresponding comprehensive evaluation function values; Each of the comprehensive evaluation function values ​​is screened, and a processing solution with the smallest comprehensive evaluation function value is selected as the optimal solution.

2. The method for generating regional power grid fault handling decisions based on knowledge graph according to claim 1 is characterized in that: The knowledge graph equivalent structure includes: The devices in the regional power grid structure are equivalent to the nodes in the knowledge graph equivalent structure, and the set of nodes is N = {n1, n2, n3, ..., n m }, m is the total number of nodes; Among them, each node has a corresponding attribute set A i ={a i1 , a i2 , a i3 , ..., a in }, A i is the attribute set of the i-th node, a i1 is the first attribute of the i-th node, a in is the nth attribute of the i-th node; The relationship between devices in the regional power grid structure is equivalent to the edges in the knowledge graph equivalent structure. The set of edges is: E = {e1, e2, e3, ..., e k }, where k is the total number of edges; Use triples to represent each edge and its two corresponding nodes. The expression is (n p , r j , n q ), where r j is the relationship type, n p is the pth node, n q is the qth node.

3. The method for generating regional power grid fault handling decisions based on knowledge graph according to claim 2 is characterized in that: Acquire the monitoring data of the regional power grid structure collected by the preset collection module, and preprocess the monitoring data to obtain a preprocessed monitoring data sequence, specifically: u monitoring points are set for the regional power grid structure, wherein the acquisition module can collect z different types of physical quantities at each monitoring point, and all the collected physical quantities constitute a monitoring data; Acquire the monitoring data collected by the acquisition module, wherein the monitoring data is represented by a u×z matrix D, D ij is the value sequence of the jth physical quantity collected at the i-th monitoring point, i = 1, 2, ..., u, j = 1, 2, ..., z, D ij (T) is the sequence of the jth physical quantity collected by the i-th monitoring point in the observation time period T, D ij (t) is the value of the jth physical quantity collected by the i-th monitoring point at time t; Based on digital filtering technology, a moving average filter is used to filter the sequence D of the jth physical quantity collected at the i-th monitoring point within time t. ij (T) is removed to obtain the filtered value Where B is the size of the moving average window; The filtered value Normalize the filtered value Normalize to the preset range, where the normalization process uses the minimum-maximum normalization process, and the formula is: in, is the sequence of the jth physical quantity collected from the ith monitoring point after preprocessing, is the minimum value of the jth physical quantity collected at the ith monitoring point during the forecast period, is the maximum value of the jth physical quantity collected by the ith monitoring point within the preset time period.

4. The method for generating regional power grid fault handling decisions based on knowledge graph according to claim 3 is characterized in that: The Pearson correlation coefficients between the plurality of preprocessed monitoring data sequences are calculated, and the fault area is preliminarily determined according to the Pearson correlation coefficient to obtain a preliminary fault area monitoring point set, specifically: Get the preprocessed monitoring data sequence and Calculate monitoring data series and Pearson correlation coefficient r iv , where v = 1, 2, ..., u, i ≠ v; Pearson correlation coefficient r is The calculation formula is: Determine the Pearson correlation coefficient r iv Whether it exceeds the preset threshold r th ; If so, it is determined that there is a related fault between the i-th monitoring point and the v-th monitoring point, and the initial fault area monitoring point set M = {i1, i2, ..., i o }, i1 is the first preliminary fault area monitoring point, i o is the oth monitoring point in the preliminary fault area, and o is the number of monitoring points in the preliminary fault area; If not, it is determined that there is no associated fault between the i-th monitoring point and the v-th monitoring point.

5. The method for generating regional power grid fault handling decisions based on knowledge graph according to claim 4 is characterized in that: Based on the knowledge graph equivalent structure, the preliminary fault area monitoring point set is searched to determine the fault area node set, the fault suspicion of each node in the fault area node set is calculated, and the most likely fault area location is determined according to the size of each fault suspicion, specifically: Search and infer the nodes corresponding to all monitoring points in the preliminary fault area monitoring point set in the knowledge graph equivalent structure to obtain the fault area node set N F ={n f1 , n f2 , ..., n fb },n f1 is the first fault region node, n fb is the bth fault area node; According to the edges of the equivalent structure of the knowledge graph and the preset electrical principle knowledge, each node in the fault area node set is calculated to calculate its fault suspicion S(n fi ), the formula is: Among them, S(n fi ) is the fault suspicion of the node in the i-th fault area, (n fi ,r j ,n fj ) is the i-th fault region node, the j-th fault region node and the relationship type r j’ The triplet of w j′ For the relationship type r j’ The importance of fault propagation and the probability of fault propagation determine the weight coefficient, I fj′ (t) is the fault judgment index corresponding to the monitoring point j′; One or more fault area nodes with the highest fault suspicion are selected from the fault suspicion of all fault area nodes, and the node is identified as the most likely fault area location. The formula is: n fault The most likely fault area location.

6. The method for generating regional power grid fault handling decisions based on knowledge graph according to claim 5 is characterized in that: Based on the preset fault pair key performance indicators of the power grid, the most likely fault area location is evaluated to determine the severity of the fault, and according to the edges of the knowledge graph equivalent structure, the set of nodes in the fault area is subjected to breadth-first search or depth-first search to obtain the affected node set, so as to determine the range of the most likely fault area location, specifically: Based on the preset fault key performance indicators of the power grid, the most likely fault area is evaluated to evaluate the fault severity. The fault severity function FS(n fi ) is: w h is the weight of the hth key performance indicator, and the key performance indicator vector for normal operation of the power grid is KPI normal ={kpi n1 ,kpi n2 ,...,kpi nr′ }, the key performance indicator vector when the fault occurs is KPI fault ={kpi f1 ,kpi f2 ,...,kpi fr′ }, r′ is the number of key performance indicator vectors; According to the edges of the knowledge graph equivalent structure, a breadth-first search or a depth-first search is performed on each fault region node in the fault region node set to obtain the affected node set IR(n fi ), specifically: Assume that the adjacency matrix of the knowledge graph equivalent structure is A1, the search depth threshold is d1, and the initialization queue Q = {n fi }, the visited node set Visited = {n fi }; When it is determined during the search process that the queue Q is not empty, the first node n of the queue Q and all nodes n' that are connected to the first node n by edges and have not been visited are extracted; Add node n' to the known queue Q and the visited node set Visited; Repeat the above steps until the search depth reaches the search depth threshold d1, and obtain the affected node set IR(n fi )=Visited.

7. The method for generating regional power grid fault handling decisions based on knowledge graph according to claim 6 is characterized in that: According to the fault severity of the most likely fault area location and the affected node set, a comprehensive evaluation function for the most likely fault area location is constructed, a plurality of processing schemes related to the most likely fault area location are extracted from a preset memory, and each processing scheme is calculated using the comprehensive evaluation function to obtain a plurality of corresponding comprehensive evaluation function values, specifically: Construct a comprehensive evaluation function G(P) for the most likely fault area location i ), wherein the comprehensive evaluation function G(P i ) is the weighted sum of the recovery time, cost and grid stability impact of the treatment solution, and the formula is G(P i )=w1t recovery (P i )+w2c(P i )+w3s stability (P i ), w1 is the weight of recovery time, w2 is the weight of cost, w3 is the weight of grid stability, w1+w2+w3=1, P i is the i-th processing solution stored in the preset memory; Among them, for each processing solution, it is assumed that it contains X operation steps, and the estimated execution time of each operation step is t iJ (J = 1, 2, ..., X), and use the directed acyclic graph DAG to represent the order of the operation steps. Let its adjacency matrix be B1. If operation step J must be performed after step K is completed, B1 JK = 1, if operation step J does not have to be performed after step K is completed, B1 JK =0; Calculate the earliest start time EST for each operation step J , EFT J =EST J +t iJ , among which, EFT J is the earliest completion time, EST1=0, EST1 is the first operation step; When J>1, EFT K is the time of the Kth operation step; Solution P i The recovery time is the earliest completion time of the last operation step, and we get t recovery (P i )=EFT X .

8. The method for generating regional power grid fault handling decisions based on knowledge graph according to claim 7 is characterized in that: Solution P i The cost c(P i ) is composed of equipment maintenance cost C repair (P i ), Human resource cost C human (P i ) and power outage loss cost C power-loss (P i )composition; Among them, suppose the set of equipment that needs to be repaired or replaced is L i = {l i1 , l i2 , ..., l iR }, the cost of repair or replacement of each device is c lj , get the equipment maintenance cost Set the number of people involved in troubleshooting to n people , each person's working hours are H, and the unit time wage is W hour , get the human resource cost C human (P i )=n people ×H×W hour ; Assume the total load in the blackout area is L load , the power outage time is t power-loss , the unit power outage loss is C loss-per-unit , and the power outage loss cost C is obtained power-loss (P i )=L load ×t power-loss ×C lpss-per-unit ; Get the treatment plan P i The cost c(P i )=C repair (P i )+C human (P i )+C power-loss (P i ).

9. The method for generating regional power grid fault handling decisions based on knowledge graph according to claim 8 is characterized in that: The grid stability impact of the treatment solution is measured by the voltage stability index VSI and the frequency stability index FSI, specifically: In the treatment plan P i After the execution is completed, the voltage stability index is obtained And frequency stability index And they are respectively the treatment scheme P i Voltage stability index And treatment plan P i Frequency stability index Assign weights w4 and w5, where w4+w5=1; Get the treatment plan P i Impact on grid stability Among them, VSI normal FSI is the voltage stability index during normal operation. normal It is the frequency stability index.

10. The method for generating regional power grid fault handling decisions based on knowledge graph according to claim 9 is characterized in that: The formula for the optimal solution is:

Images