Power grid maintenance risk intelligent assessment and dynamic prevention and control strategy generation method

Abstract

The invention relates to the technical field of power grid maintenance, in particular to a power grid maintenance risk intelligent assessment and dynamic prevention and control strategy generation method, which comprises the following steps of: constructing a power grid CIM model based on a knowledge graph; performing multi-source data fusion and feature extraction; analyzing the power grid topology; checking the maintenance constraint rule; intelligent risk assessment; and generating a dynamic prevention and control strategy. According to the method, the power grid CIM model based on the knowledge graph is constructed, equipment, lines and the like are stored in a graph structure, efficient expression and retrieval of the power grid topological relation are achieved, the LSTM multivariate time sequence model integrated with multiple variables is adopted, load characteristics are analyzed and predicted, abnormal values and periodic characteristics in historical load data are effectively processed, and the reliability of the power grid topological relation is improved. And multi-source data fusion and intelligent analysis are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power grid maintenance, in particular to a power grid maintenance risk intelligent evaluation and dynamic prevention and control strategy generation method. BACKGROUND

[0002] In the operation of the power system, power grid maintenance is a key link to ensure power supply reliability. However, the traditional power grid maintenance risk evaluation and prevention and control strategy has many shortcomings. First, the traditional method is difficult to effectively integrate maintenance plans, real-time operation data, equipment defect libraries and historical risk data and other multi-source heterogeneous data. The analysis of load characteristics is mostly based on single variable models, which cannot accurately capture the complex rules of power load changes with time, environment and other factors, resulting in insufficient load prediction accuracy. Second, the analysis of power grid topology structure stays at the static modeling level, lacking a dynamic search mechanism for main lines, loops and branch nodes. It is difficult to quickly evaluate the transferable load path and power loss impact range during equipment maintenance, resulting in a long period of maintenance scheme development. In addition, the generation of maintenance strategies lacks multi-objective optimization. The development of maintenance plans relies on manual experience, and it is difficult to achieve global optimization among safety constraints, resource limitations and economic targets, often leading to a balance between maintenance efficiency and power supply reliability.

[0003] The invention patent with publication number CN117057590A provides a power grid maintenance management system and method, which includes: a power grid maintenance task application module detects the published power grid maintenance work, extracts the power grid maintenance content and power grid outage plan, generates and submits a power grid maintenance task application; a power grid maintenance risk analysis module identifies the adjustment of power grid operation mode during maintenance, analyzes the power grid maintenance risk of each station, and generates an accident handling plan by referring to the power grid maintenance fault handling library; a power grid maintenance task issuing module generates a power grid maintenance risk warning notice according to the accident handling plan of the power grid maintenance task approved by the superior power grid department, issues it to the operation and maintenance personnel, and guides the power grid maintenance work; a power grid maintenance fault handling library module saves the actual solution and accident solving effect when a power grid accident occurs during power grid maintenance.

[0004] Although the application standardizes the process of power grid maintenance task application, recommends accident handling plan, and guarantees the efficient power grid maintenance of operation and maintenance personnel, there are still many deficiencies, one of which is that in the aspect of multi-source data fusion and load characteristic analysis, the traditional method is difficult to effectively fuse the multi-source heterogeneous data such as maintenance plan, real-time operation data, equipment defect library and historical risk data, and the analysis of load characteristics is mostly based on a single variable model, which cannot accurately capture the complex rules of power load changes with time, environment and other factors, resulting in insufficient load prediction accuracy; the second is that in the aspect of power grid topology analysis, the analysis of power grid topology structure stays at the static modeling level, lacks dynamic search mechanism of main line, loop and branch node, and it is difficult to quickly evaluate the transferable load path and power loss influence range during equipment maintenance, resulting in long period of maintenance scheme formulation; the third is that in the maintenance strategy generation link, there is lack of multi-objective optimization, and the maintenance plan is formulated depending on artificial experience, which is difficult to achieve global optimization among safety constraints, resource limitations and economic targets, often leading to imbalance between maintenance efficiency and power supply reliability; therefore, the application provides a power grid maintenance risk intelligent evaluation and dynamic prevention and control strategy generation method. SUMMARY

[0005] The application aims to provide a power grid maintenance risk intelligent evaluation and dynamic prevention and control strategy generation method to solve the problems in the background art.

[0006] To solve the above technical problems, the application provides a power grid maintenance risk intelligent evaluation and dynamic prevention and control strategy generation method, which comprises the following steps:

[0007] S1, a power grid CIM model based on a knowledge graph is constructed, CIM model files of a main grid and a distribution network are acquired, network topology relationships are generated by analyzing the CIM model files, devices, lines and electrical equipment containers are taken as graph vertices, relationships are taken as edges connecting the vertices, and the graph is stored in the form of a graph library and stored in a graph database;

[0008] S2, multi-source data fusion and feature extraction, a feature model is constructed, maintenance plans, real-time operation data, equipment defect libraries and historical risk data are collected, a multivariate LSTM multivariate time sequence is adopted for load characteristic analysis and prediction, a load characteristic feature model is constructed, important users and large users are considered to construct a load distribution feature model, and a power supply capacity feature model, a maintenance capacity feature model and a maintenance capacity feature model are constructed;

[0009] S3, power grid topology analysis, deep-first search is adopted to find main lines and loops in the power grid, the hierarchical relationship and connection mode between lines are determined, breadth-first search is adopted to find branches and terminal nodes in the power grid, the branch structure and hierarchical relationship of the power grid are established, and transferable load search, N-1 criterion analysis and power loss equipment analysis are performed;

[0010] S4, overhaul constraint rule checking, the checking of the overhaul constraint rules includes safety constraints, coordination constraints, overhaul resource constraints, equipment constraints and economic constraints, and a correlation is established by using a knowledge graph technology;

[0011] S5, risk intelligent assessment, based on a power grid CIM model and power grid topology analysis, in combination with equipment health status and weather influencing factors, an overhaul risk level is assessed, and a key equipment health degree is identified;

[0012] S6, dynamic prevention and control strategy generation, a genetic algorithm is used to perform multi-objective optimization on the overhaul plan, and a risk degradation scheme and a prevention and control measure are generated.

[0013] As a further improvement of the technical solution, in S1, when the knowledge graph model is constructed, pruning optimization is performed, the endpoints are trimmed, the connection points are directly connected by the equipment, the topology path is shortened, and the number of graph library nodes is reduced.

[0014] This setting shortens the topology path and reduces the number of graph library nodes through pruning optimization, which can significantly improve the query efficiency and topology analysis speed of the graph database, reduce the consumption of computing resources, and at the same time simplify the complexity of the power grid model, so that the topology relationship is expressed more simply, and the subsequent risk assessment and strategy generation are efficiently executed.

[0015] As a further improvement of the technical solution, in S2, a multivariate LSTM multivariate time series is used for load characteristic analysis and prediction, and the specific steps are as follows:

[0016] A1, collect historical power load data, divide it into multiple time steps according to time sequence, and input data format is time sequence vector, the shape of each time step data is (batch_size, input_dim), wherein batch_size is batch size, and input_dim is feature dimension

[0017] A2, remove abnormal load data, and normalize the data;

[0018] A3, construct the core structure of the LSTM model, the input content includes cell state C t-1 , hidden layer state h t-1 and input vector x t at time t, output cell state C t and hidden layer state h t at next time, and the cell state and the hidden layer state are initialized as all-zero vectors, and there are three gate structures of forget gate, input gate and output gate in each neural network layer to protect and control information, the forget gate is used to filter historical information, the input gate is used to generate candidate state and update the cell state, and the output gate is used to determine the current output, and the corresponding formula is:

[0019] f t = sigma(w f [ h t-1 , x t ] + b t )

[0020] i t = sigma(w i [ h t-1 , x t ] + b i )

[0021]

[0022]

[0023] o t = sigma(w o [ h t-1 , x t ] + b o )

[0024] h t = o t tanh(C t )

[0025] In the formula, f t represents the output of the forget gate, sigma represents the sigmoid function, w f represents the forget gate weight, b f represents the forget gate bias, [h t-1 , x t ] represents the connection of two vectors, i t represents the output of the input gate, w i represents the input gate weight, b i represents the input gate bias, represents the candidate cell state, tanh represents the activation function, w c represents the calculation weight of the candidate cell state, b c represents the calculation bias of the candidate cell state, o t represents the output of the output gate, w o represents the output gate weight, b o represents the output gate bias.

[0026] As a further improvement of the technical solution, in S2, a power supply capacity feature model is constructed based on the maximum power supply capacity, the maximum load supply capacity of the feeder is calculated according to the capacity of the main transformer and the feeder in the power grid and the topological structure of the network, and the main transformer power supply capacity, feeder comprehensive current, line bottleneck current, CT transformation ratio and highest current in the last three months are comprehensively evaluated:

[0027]

[0028] In the formula, S max represents the maximum power of the feeder, S 3,max represents the maximum power of the feeder in the last three months, S x represents the power that needs to be transferred, I max represents the maximum power supply current of the feeder, I 3,max represents the highest current of the feeder in the last three months, I x represents the current that needs to be transferred, S B represents the power supply capacity of the main transformer, I lc represents the comprehensive current-carrying capacity of the feeder, represents the maximum current-carrying capacity allowed by the wire diameter, I ct represents the CT transformation ratio of the feeder, and min(·) represents taking the minimum value;

[0029] The power supply capacity of the main transformer S B depends on the capacity of the main transformer in the substation, and the calculation formula is:

[0030]

[0031] In the formula, S in represents the rated capacity of the high-voltage side of the transformer, S out represents the rated capacity of the low-voltage side of the transformer, represents the power factor;

[0032] The CT transformation ratio is the ratio between the current transformation on both sides of the current transformer.

[0033] As a further improvement of the technical solution, in the S2, the specific steps for constructing the maintenance capacity feature model are:

[0034] B1, collect the number of power-off equipment, the number of safety protection groups, the number of distribution automation switches, and the operation time of each department in the power grid maintenance record to form a structured data set;

[0035] B2, clean the data, unify the time unit, classify the equipment types, and eliminate outliers;

[0036] B3, determine the input variables and output variables, the input variables X include the number of power-off equipment X1, the number of safety protection groups X2, and the number of distribution automation switches X3, and the output variables Y include the installation time of station safety protection Y1, the removal time of station safety protection Y2, the power-off operation time Y3, the power-on operation time Y4, the power-off order time Y5, and the power-on order time Y6;

[0037] B4, establish a linear regression model, and use a multiple linear regression equation Y = β0 + β1X1 + β2X2 + β3X3 + ε, wherein β0 is the intercept, β1, β2, and β3 are regression coefficients, and ε is an error term.

[0038] B5, using the least square method to solve the coefficient, fitting the parameters through the historical data, making the error square sum of the predicted value and the actual value minimum.

[0039] As a further improvement of the technical solution, in S2, the maintenance capacity characteristic model is constructed, the ratio of the reserved standby capacity of the online operation power station participating in power generation dispatching to the maximum working capacity is equal to the ratio of the total standby capacity of the power system to the annual maximum load of the power system, the sum of the reserved standby capacity of the power station is equal to the total standby capacity of the power system, that is:

[0040]

[0041] In the formula, S A represents the standby capacity reserved by the online power station, S represents the maximum working capacity of the system, S D represents the total standby capacity of the system, S M represents the annual maximum load of the system.

[0042] Among the four settings, the multi-variable LSTM model is used to realize accurate analysis of load characteristics, effectively handle historical data outliers and periodic characteristics, improve load prediction accuracy, and provide reliable data support for maintenance plan making; the power supply capacity characteristic model is constructed to evaluate the maximum load supply capacity of the feeder from multiple dimensions to ensure the power supply reliability and stability of the power grid during maintenance; the maintenance capacity characteristic model is fitted through historical data to realize accurate prediction of the maintenance operation time, help reasonably arrange the maintenance process, and improve the maintenance efficiency; the maintenance capacity characteristic model is based on the standby capacity proportion relationship to ensure the power supply redundancy of the system during maintenance, balance the power generation dispatching and maintenance demand, and improve the safety of power grid operation.

[0043] As a further improvement of the technical solution, in S3, the transferable load search, N-1 criterion analysis and power failure equipment analysis are performed, the transferable load search is calculated by graph traversal of the power failure equipment, the traversal is performed by depth-first algorithm, the available power supply point is traced back, the N-1 criterion analysis is analyzed and traversed by the minimum spanning tree algorithm, the power failure equipment analysis is traversed by the minimum spanning tree algorithm to form a device set after failure of the current running device, and the complement set calculation is performed on the two sets to obtain the current power failure equipment set.

[0044] In this setting, the transferable load search quickly traces back the power supply point by the depth-first algorithm to provide the transfer scheme for the power failure equipment, reduce the power failure range and time, and ensure the power supply continuity; the N-1 criterion analysis ensures the safety of the power grid when a single component fails by means of the minimum spanning tree algorithm, and improves the risk resistance of the system; the power failure equipment analysis accurately locates the failure influence range by set complement calculation to provide clear targets for risk assessment and prevention strategies, and improve the accuracy of maintenance decision.

[0045] As a further improvement of the technical solution, in the S4, the checking of the maintenance constraint rules includes safety constraints, coordination constraints, maintenance resource constraints, equipment constraints and economic constraints, the safety constraints include active power flow constraints, transfer capacity constraints and protection action verification, the coordination constraints include simultaneous maintenance constraints, sequential maintenance constraints, mutually exclusive maintenance constraints and unchangeable maintenance constraints, the maintenance resource constraints include maintenance unit constraints and dispatching resource constraints, the equipment constraints include power transmission and transformation combination constraints and maximum number of maintenance tie lines, and the economic constraints include reducing load loss and reducing power outage time.

[0046] In this setting, the multi-dimensional maintenance constraint rules are checked through the knowledge graph association to realize the automatic checking of safety, coordination, resource, equipment and economic constraints, avoiding safety hazards and resource waste caused by manual omission; the safety constraints ensure the rationality of power flow distribution and protection action during maintenance, preventing equipment overload and misoperation; the coordination constraints optimize the maintenance sequence and combination, reducing repeated power outage and maintenance conflicts and improving maintenance efficiency; the resource and equipment constraints reasonably allocate maintenance resources, avoiding excessive maintenance or insufficient resources and reducing maintenance cost; and the economic constraints balance maintenance investment and power supply benefit by reducing load loss and power outage time, improving the economic efficiency of power grid operation.

[0047] As a further improvement of the technical solution, in the S5, based on the power grid CIM model and power grid topology analysis, combined with the health status of equipment and weather influencing factors, the maintenance risk level is evaluated, and the key equipment health degree is identified, and the specific steps are as follows:

[0048] C1, calculate the real-time failure rate of the equipment according to the health index of the equipment state;

[0049] C2, sequentially evaluate the reduced load, identify the voltage loss plant, evaluate the influence on users, assess the event level and risk level.

[0050] As a further improvement of the technical solution, in the S6, the dynamic prevention and control strategy is generated, the genetic algorithm is used for multi-objective optimization of the maintenance plan, and the risk degradation scheme and prevention and control measures are generated, and the specific steps are as follows:

[0051] D1, number the maintenance plan, slice by time, fix the information in the given time table, create a chromosome, randomly allocate time slices, generate a plan maintenance single chromosome and encode;

[0052] D2, establish a fitness function, input the plan maintenance single information and time slice, after verification and optimization, return the fitness of the maintenance plan;

[0053] D3, perform selection, crossover and mutation operations to generate a new population;

[0054] D4, algorithm termination judgment.

[0055] In the two settings, the risk intelligent assessment combines the device health state and weather factors, realizes the dynamic assessment of the maintenance risk level through the quantitative relationship between the health index and the real-time failure rate, provides accurate basis for risk prevention and control, the dynamic prevention and control strategy adopts the genetic algorithm for multi-objective optimization, generates the maintenance scheme considering the safety constraints, resource limitations and economic targets, realizes the intelligent generation of risk degradation and prevention and control measures, and improves the global optimality and reliability of power grid maintenance.

[0056] Compared with the prior art, the beneficial effects of the present application are:

[0057] 1. In the power grid maintenance risk intelligent assessment and dynamic prevention and control strategy generation method, the power grid CIM model based on the knowledge graph is constructed, the devices, lines and the like are stored in a graph structure, efficient expression and retrieval of the power grid topological relationship are realized, the LSTM multivariate time series model is adopted to analyze and predict the load characteristics, the abnormal values and periodic characteristics in the historical load data are effectively processed through the dynamic control of the forgetting gate, the input gate and the output gate, and multi-source data fusion and intelligent analysis are realized.

[0058] 2. In the power grid maintenance risk intelligent assessment and dynamic prevention and control strategy generation method, the depth-first search and breadth-first search algorithms are combined to realize the rapid identification of the main line, loop and branch structure of the power grid, based on the device health state and weather influence factors, the quantitative relationship between the health index and the real-time failure rate is realized to dynamically assess the maintenance risk level, accurate basis is provided for risk prevention and control, and the accuracy of topological analysis and risk assessment is realized.

[0059] 3. In the power grid maintenance risk intelligent assessment and dynamic prevention and control strategy generation method, the multi-dimensional maintenance constraint rules are associated and verified through the knowledge graph, avoiding safety hazards caused by manual missed detection, the maintenance scheme generated by the genetic algorithm multi-objective optimization simultaneously satisfies the maintenance resource and economic constraints, and the balance between safety and benefit is realized. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 The method flowchart of the present application. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0062] AsFigure 1 As shown, the embodiment provides a power grid maintenance risk intelligent assessment and dynamic prevention strategy generation method, including the following steps:

[0063] S1, a knowledge graph-based power grid CIM model is constructed, CIM model files of the main grid and distribution network are acquired, network topology relationships are generated by parsing the CIM model files, devices, lines, and electrical equipment containers are used as graph vertices, relationships are used as edges connecting the vertices, and the graph is stored in the graph database in the form of a graph library;

[0064] After parsing the CIM model file, busbar, station, main transformer, line, disconnector, and grounding switch device data and relationships can be obtained. The CIM model is an abstract model that describes objects related to power operation, which is defined by multiple logical packages. The knowledge graph is a structured semantic knowledge base that can describe physical objects and relationships of power grid control information in a visual form using graphical symbols. Its basic unit is a triple of "entity-relation-entity", "entity-attribute-value". Entities are connected to each other through relationships, forming a network-like knowledge structure. The knowledge graph describes concepts, entities, events, and their complex relationships in the objective world in a structured form. Concepts refer to the conceptualized representation of objective things in the process of understanding the world, such as controllers, maintenance personnel, line patrol personnel, substations, lines, and towers. Entities are specific things in the objective world, such as XX power station, XX power plant, and controllers. Events are activities of objective events, such as fault handling and maintenance. Relationships describe the association between concepts, entities, and events, such as the relationship between switches, disconnectors, and other devices within a substation, the relationship between switches and transformers, and the relationship between three-winding transformers. By effectively integrating power grid basic information, operation data, and dispatching knowledge, a knowledge graph for online analysis and decision-making support in power grid control can be constructed, which facilitates fast retrieval, reasoning, and analysis of risks, abnormalities, faults, voltage, and load, and provides intelligent decision support.

[0065] A graph database is a database that uses a graph structure for semantic queries. It uses points, edges, and attributes to represent and store data. The expressive power of a graph database is enhanced, and it can describe the data form in real-world scenarios. The core goal of graph database storage is to achieve index-free adjacency. This storage method has the characteristics of graph element association analysis and good support for power grid topology storage and application. It supports visual display and maintenance management of data logical relationships and dynamic creation of relationships between data, making it suitable for storing large-scale and complex knowledge graphs.

[0066] S2, multi-source data fusion and feature extraction, construct a feature model, collect maintenance plans, real-time operation data, device defect library and historical risk data, use LSTM multivariate time series to analyze and predict load characteristics, construct a load characteristic feature model, consider important users and large users to construct a load distribution feature model, construct a power supply capacity feature model, a maintenance capacity feature model and a maintenance capacity feature model;

[0067] S3, power grid topology analysis, use depth-first search to find the main line and loop in the power grid, determine the hierarchical relationship and connection mode between lines, use breadth-first search to find branches and terminal nodes in the power grid, establish the branch structure and hierarchical relationship of the power grid, and perform load transfer search, N-1 criterion analysis and loss of equipment analysis;

[0068] S4, maintenance constraint rule checking, the checking of maintenance constraint rules includes safety constraints, coordination constraints, maintenance resource constraints, device constraints and economic constraints, and the correlation is established using knowledge graph technology;

[0069] S5, risk intelligent assessment, based on power grid CIM model and power grid topology analysis, combined with device health status and weather influence factors, assesses the maintenance risk level and identifies the key device health degree;

[0070] S6, dynamic prevention and control strategy generation, uses genetic algorithm for multi-objective optimization of maintenance plan, generates risk reduction scheme and prevention and control measures.

[0071] In the embodiment, in S1, when constructing the knowledge graph model, the device, line, and electrical device container are used as the graph vertex, and the relationship is used as the edge connecting the vertex to construct. Similarly, the dependency relationship and connection relationship between electrical devices are also constructed. If the CIM model is completely modeled, it is generally composed of "Device 1-Endpoint 1-Connection Point-Endpoint 2-Device 2". The endpoint and connection point clearly describe the connection relationship between electrical devices. However, constructing a knowledge graph model according to the above structure will directly lead to a longer path during knowledge graph topology derivation, and also increase the number of nodes, affecting the performance of graph calculation. To improve efficiency and reduce calculation, pruning optimization is performed on the endpoints, and the device is directly connected to the connection point, shortening the topology path and reducing the number of graph library nodes.

[0072] Specifically, in S2, the load characteristics of the power system are the rules of active power and reactive power drawn by the power load from the power source of the power system changing with the voltage at the load end and the system frequency. Load characteristics are an important part of the power system, and they have an important influence on the analysis, design and control of the power system as consumers of electric energy. LSTM multivariate time series are used to analyze and predict load characteristics, and the specific steps are as follows:

[0073] A1, collect historical power load data, divided into multiple time steps according to time sequence, input data format is time sequence vector, shape of each time step data is (batch_size, input_dim), wherein batch_size is batch size, and input_dim is feature dimension

[0074] A2, remove abnormal load data, and normalize the data

[0075] A3, construct the core structure of the LSTM model, and the input content includes cell state C t-1 , hidden layer state h t-1 and input vector x t at time t, output cell state C t and hidden layer state h t at next time, the cell state and the hidden layer state are initialized as all-zero vectors, and there are three gate structures of forget gate, input gate and output gate in each neural network layer to protect and control information, the forget gate is used to filter historical information, the input gate is used to generate candidate state and update the cell state, and the output gate is used to determine the current output, and the corresponding formula is:

[0076] The forget gate filters information and decides which information can pass through the unit, the load characteristics are predicted by historical load data, at this time, the state of each information storage unit contains the load information of the current period, but the single prediction time contains weekends and holidays, and it is necessary to forget such information because the load of weekends and holidays is very different from that of normal days, and the load mode should be used in the same period:

[0077] f t =σ(w f [h t-1 ,x t ]+b f )

[0078] The input gate updates the information state of the unit, and in the prediction of the load characteristics, new information is input into the unit state, and the old state information is replaced:

[0079] i t =σ(w i [h t-1 ,x t ]+b i )

[0080]

[0081]

[0082] The output gate needs to determine the input content, in the load feature analysis prediction, assuming that the model just contacts the load information of a day, and the load mode of the day is output, it is necessary to input the information learned about the day to the unit state, so as to carry out the next step of prediction:

[0083] o t =σ(w o ) t-1 , x t ]+b o )

[0084] h t =o t tanh(C t )

[0085] In the formula, f t represents the output of the forgetting gate, sigma represents the sigmoid function, w f represents the forgetting gate weight, b f represents the forgetting gate bias, [h t-1 , x t ] represents the connection of two vectors, i t represents the output of the input gate, w i represents the input gate weight, b i represents the input gate bias, represents the candidate cell state, tanh represents the activation function, w c represents the calculation weight of the candidate cell state, b c represents the calculation bias of the candidate cell state, o t represents the output of the output gate, w o represents the output gate weight, b o represents the output gate bias.

[0086] In S2, important users and large users are considered to construct a load distribution feature model. With the continuous expansion of the power grid and the increasing complexity of the structure, power grid maintenance also becomes more and more complex. Unreasonable maintenance will cause user power outage to the power grid, especially to important power users and large users. The power outage not only causes economic losses, but also may cause certain environmental pollution, personal accidents and other important social losses such as serious social losses. Therefore, when arranging the power grid maintenance, the influence of maintenance on important users and large users should be fully considered. For important users and large users, the load distribution feature model specific to important users and large users is diagnosed by using the load feature analysis model, and in the maintenance process, power outage to important users and large users is avoided as much as possible. If it is really impossible to avoid, multiple power supply modes should be adopted to ensure the stability and reliability of power supply.

[0087] Further, in the S2, a power supply capability feature model is constructed based on maximum power supply capability, which refers to the maximum load that the power distribution network can carry when all feeder N-1 checks and transformer N-1 checks of the transformer substation in the power distribution network are satisfied. The load transfer between the transformer and the feeder, the capacity of the transformer and the feeder, the connection relationship between the transformers and between the feeders in the network, and other actual operation constraints of the power distribution network are considered. According to the capacity of the transformer and the feeder in the network and the topology of the network, the maximum load supply capability of the feeder is calculated, and the maximum power supply capability of the transformer, the comprehensive current of the feeder, the line bottleneck current, the CT transformation ratio, and the maximum current of the feeder in the last three months are comprehensively evaluated:

[0088]

[0089] In the formula, S max represents the maximum power of the feeder, S 3,max represents the maximum power of the feeder in the last three months, S x represents the power that needs to be transferred, I max represents the maximum power supply current of the feeder, I 3,max represents the maximum current of the feeder in the last three months, I x represents the current that needs to be transferred, S B represents the power supply capability of the transformer, I lc represents the comprehensive current carrying capacity of the feeder, represents the maximum current carrying capacity allowed by the line diameter, I ct represents the CT transformation ratio of the feeder, and min(·) represents the minimum value;

[0090] The power supply capability of the transformer S B depends on the capacity of the transformer in the transformer substation. When analyzing the power supply capability of the transformer, the size of the load carried by the transformer is calculated, and the formula is as follows:

[0091]

[0092] In the formula, S in represents the rated capacity of the high-voltage side of the transformer, S out represents the rated capacity of the low-voltage side of the transformer, represents the power factor;

[0093] The comprehensive current carrying capacity of the feeder I lc is the comprehensive evaluation value of the maximum load that the feeder can carry;

[0094] The maximum current carrying capacity allowed by the line diameter There are two kinds, one is due to the influence of physical limit, the minimum line segment allowed by the current in the feeder, the other is to consider the distribution and operation of the load, to evaluate the whole feeder bottleneck; For the former, the minimum bottleneck can be calculated according to the line diameter parameters of all lines in the feeder, and for the latter, because the load is randomly dispersed and the operation law is uncertain, it is difficult to capture the load characteristics. For this situation, the load characteristics model mentioned above is used to predict the load on the feeder, and the line bottleneck is dynamically evaluated based on the load condition;

[0095] The CT transformation ratio is the ratio between the currents on both sides of the current transformer, that is, the ratio of the current size on the primary side and the secondary side. The primary side is the high-voltage part of the line, and the secondary side is the low-voltage part to be processed. It is mainly used as a current measurement element in the current loop of the secondary circuit for metering, measurement, relay protection, monitoring, etc. In the power supply characteristic analysis, the CT transformation ratio needs to be considered to avoid causing protection action.

[0096] Further, in S2, the specific steps of constructing the maintenance capability feature model are:

[0097] B1, collect the number of power outage equipment, safety measure group number, distribution automation switch number, and each department operation time in the power grid maintenance record to form a structured data set;

[0098] B2, clean the data, unify the time unit, classify the equipment types, and eliminate outliers;

[0099] B3, determine the input variables and output variables, the input variables X include the number of power outage equipment X1, the number of safety measure groups X2, and the number of distribution automation switches X3, and the output variables Y include the installation time of station safety measures Y1, the removal time of station safety measures Y2, the power outage operation time Y3, the power restoration operation time Y4, the power outage order time Y5, and the power restoration order time Y6;

[0100] B4, establish a linear regression model, use a multiple linear regression equation Y = β0 + β1X1 + β2X2 + β3X3 + ε, where β0 is the intercept, β1, β2 and β3 are regression coefficients, and ε is the error term;

[0101] For example, to calculate the influence of the number of power outage equipment X1, the number of safety measure groups X2, and the number of distribution automation switches X3 on the installation time of station safety measures Y1, the equation established is Y1 = β0 + β1X1 + β2X2 + β3X3 + ε;

[0102] B5, use the least squares method to solve the coefficients, and fit the parameters through historical data to minimize the error sum of squares of the predicted value and the actual value.

[0103] Further, in the S2, a maintenance capacity characteristic model is constructed. The maintenance capacity refers to the capacity of a device that is maintained or repaired according to a plan in the power system. When a generator unit in the system is repaired, it is necessary to consider whether the capacity of the repaired unit is less than the standby capacity of the power grid, i.e., the equal standby principle. The ratio of the standby capacity reserved by an online operating power station participating in power generation dispatch to the maximum working capacity in the power system is equal to the ratio of the total standby capacity of the system to the annual maximum load of the power system. The sum of the standby capacities reserved by the power stations is equal to the total standby capacity of the power system, i.e.,

[0104]

[0105] In the formula, S A represents the standby capacity reserved by an online power station, S represents the maximum working capacity of the system, S D represents the total standby capacity of the system, S M represents the annual maximum load of the system.

[0106] Further, in the S3, a transferable load search, N-1 criterion analysis, and power loss equipment analysis are performed.

[0107] The transferable load search is calculated by graph traversal of the power loss equipment, and the depth-first algorithm is used for traversal. The power loss equipment is traced to the available power supply point. The power loss equipment that can be traced to one or more power supply points is the transferable load. The main function of the power supply tracing is to automatically trace the power supply point of the in-station equipment or out-station feeder equipment to the power supply point of the equipment through the network topology tracing function, and to display the power supply path and the power supply point on the power grid connection diagram. The power supply point is defined as the equivalent power supply of a substation. The in-station equipment can be a transformer, bus, switch, load, capacitor, etc. The out-station equipment can be a feeder segment, transformer, bus, switch, load, capacitor, etc. The power supply tracing starts from one end of the traced equipment and performs a depth-first search. When the power supply is found, the power supply tracing is stopped. If the power supply is found, the power supply tracing is ended, and the power supply and the power supply path are colored and displayed on the network. Otherwise, the depth-first search is performed from the other side of the traced equipment. When the power supply is found, the power supply tracing is stopped. If the power supply is found, the power supply tracing is ended, and the power supply and the power supply path are colored and displayed on the network. Otherwise, the search for the power supply fails. When the traced equipment is a bus, load, or capacitor, it is a single-end equipment, and thus only one side needs to be searched. When the traced equipment is a switch or feeder segment, it is a double-end equipment, and thus both sides need to be searched. When the traced equipment is a transformer, the transformer supplies power from the high-voltage side, and thus only the high-voltage side needs to be searched.

[0108] The depth-first search algorithm is implemented by using two data structures, a linked list and a stack. The linked list is used to store power supply sources and power supply paths. The stack is used to store all devices searched in the depth-first search process and adopts a last-in first-out strategy. In the depth-first search, a tracking device is first in-stacked, then out-stacked and added to the linked list. Devices directly connected to the tracking device are searched and in-stacked. The in-stacked devices are out-stacked and added to the linked list. Devices connected to the out-stacked devices but not in-stacked are searched and in-stacked. The process is repeated until the power supply device is found, that is, the search of a power supply path is completed.

[0109] The N-1 criterion analysis is performed by using the minimum spanning tree algorithm to analyze and traverse the connection relationship. The N-1 criterion contains two meanings: one is to ensure the stability of the power grid; and the other is to ensure that the user obtains continuous power supply meeting the quality requirements. The N-1 criterion is used for static security analysis of the power system under the condition of single-element fault disconnection, or dynamic security analysis of the power system after single-element fault disconnection. Compared with the reliability analysis, the N-1 criterion is simple to calculate and does not need to collect a large amount of original data such as element outage rate, and is a very simple security check criterion. According to the knowledge graph model constructed above, the topological deduction is performed to analyze the loss of power devices or search for transfer paths, which is essentially a path search on the graph data. The minimum spanning tree algorithm is used to analyze and traverse the connection relationship.

[0110] The loss of power device analysis is performed by using the minimum spanning tree algorithm to traverse the devices in the running state to form a set of devices after fault. The current loss of power device set is obtained by calculating the complement of the two sets.

[0111] In addition, in the S4, the checking of the maintenance constraint rules includes safety constraints, coordination constraints, maintenance resource constraints, device constraints and economic constraints. The safety constraints include active power flow constraints, transfer capacity constraints and protection action verification. The coordination constraints include simultaneous maintenance constraints, sequential maintenance constraints, mutually exclusive maintenance constraints and non-changeable maintenance constraints. The maintenance resource constraints include maintenance unit constraints and dispatching resource constraints. The device constraints include power transmission and transformation combination constraints and maximum maintenance tie-line number. The economic constraints include reduction of load loss and reduction of power outage time. Specifically,

[0112] The active power flow constraint is to ensure the safe and normal operation of the power grid equipment under the N-1 criterion. When the equipment is maintained, the system power flow distribution changes under the N-1 criterion, which may cause the transmission power of the line to exceed the limit. Therefore, the line power flow must be calculated and verified.

[0113] The transfer capacity constraint is to coordinate the main grid transfer capacity in the case of power failure area transfer. In order to ensure power supply reliability, during equipment maintenance, it is often necessary to consider the transfer of power failure users. When transferring, the power supply capacity needs to be evaluated to ensure that the power supply equipment has sufficient power supply capacity.

[0114] The protection action verification is to verify the protection setting value when the maintenance causes the operation mode to change. Equipment maintenance will inevitably change the operation mode of the power grid. Different operation modes affect the scope of protection and may cause the operation protection action to trip. After the operation mode changes, the protection action must be verified.

[0115] Simultaneous maintenance constraint is the simultaneous maintenance of interval equipment with the same logic and the maintenance of the main grid. In a system, power outage maintenance plan arrangement is very important. The problems that can be solved by one-time power outage determine that multiple power outage maintenance is not allowed, thereby causing repeated power outage problems, such as line and two-side switch, main transformer and three-side switch, 110kV main transformer and 10kV bus, T-connected main transformer corresponding line, line transformer group corresponding line, single bus corresponding line and main transformer, bus and bus section switch, all switches on the bus, etc. All maintenance on the power failure node should be combined together and unified in time for maintenance.

[0116] Sequential maintenance constraint is to perform sequential maintenance according to the constraints of the equipment. Due to the large geographical area covered by the power supply system, the number of maintenance personnel and the amount of maintenance funds are limited. When formulating the maintenance plan, the rationality of the maintenance route of the maintenance personnel should be fully considered, and the maintenance sequence should be reasonably arranged according to the principle of proximity to the geographical location, so as to reduce labor intensity and maintenance management costs such as travel expenses, thereby improving economic efficiency.

[0117] Mutually exclusive maintenance constraint is to perform mutually exclusive verification on double-circuit lines, main transformers in the same station, and other equipment to avoid forming an electrical island. In order to avoid load outage during maintenance, devices that are mutually reserved cannot be maintained at the same time, such as equipment in the same power plant, outgoing or incoming lines in the same station 110kV, main transformers in the same station, busbars of the same voltage level in the same station, PTs on parallel buses, double-circuit lines, etc. They cannot be arranged for maintenance at the same time during maintenance.

[0118] Non-changeable maintenance constraint is to directly include the upper-level designated, last month's remaining, fault-generated, and designated maintenance without constraint verification.

[0119] Maintenance unit constraint is due to the limitation of the number of personnel and technical ability, so that the number of maintenance equipment is limited.

[0120] Dispatching resource constraint is the ability of dispatching personnel to perform power outage and restoration operations in a single day. Due to the limited number of dispatching personnel and the heavy workload, there is a certain limit to the number of power outage and restoration in a single day according to different dispatching execution capabilities.

[0121] The transmission and transformation combination constraint is that for a special transmission and transformation combination, the combination constraint should be checked according to the special combination mode, and the combination constraint condition should be met during maintenance;

[0122] The maximum number of maintenance contact lines is to guarantee the power supply capacity of the power grid, and the number of contact lines that can be maintained by the distribution network is limited;

[0123] Reducing load loss is to arrange maintenance during the low load period;

[0124] Reducing the length of power outage

[0125] In order to reduce the loss of electricity, reasonably arrange the maintenance time, reduce the loss of power outage, and adjust the maintenance time.

[0126] It is worth noting that in S5, based on the power grid CIM model and power grid topology analysis, combined with the health state of the equipment and weather influencing factors, the maintenance risk level is evaluated, and the key equipment health degree is identified, and the specific steps are:

[0127] C1, the real-time failure rate of the equipment is calculated according to the health index of the equipment state, the quantitative relationship between the health index H and the real-time failure rate λ of the equipment is described as:

[0128] λ=Kexp(CH)

[0129] In the formula, K is the proportional coefficient, and C is the curvature coefficient;

[0130] C2, the reduced load, the identification of voltage loss power station, the evaluation of the influence of users, the evaluation of the event level and the risk level assessment are carried out in turn;

[0131] The reduced load evaluation is to simulate the fault and evaluate the load cut off by the fault, and the calculation method is to take the active power of the load line and use the value of the base section to evaluate. The loss of load calculation method is as follows:

[0132]

[0133] In the formula, P loss is the reduced load for evaluation, P i1 is the load of each power loss equipment, and P j2 is the load of each transferable equipment;

[0134] When evaluating the reduced load, the active power of each main device in the base state section needs to be stored as part of the model, and the maximum load in the recently collected load data is usually taken as the load of the device for evaluation. If the load transfer occurs during the time period, the maximum load will include the transferred part, so the load cannot accurately reflect the actual reduced load. Therefore, when using recent data, the part containing the transfer needs to be removed, and the base operation mode is identified by the data similarity analysis method. The active power of all distribution transformers in the past three months is analyzed for similarity analysis, and the outliers are found and removed from the data in the time range. Then, statistics are made;

[0135] The identification of the voltage loss station is automatically analyzed by the outage equipment in the maintenance mode. After searching for the outage equipment through the power supply range analysis, the voltage loss of the station to which the outage equipment belongs can be identified. The station voltage loss / operation state is analyzed under the current power grid state. The identification method can isolate the voltage loss equipment and trace the power supply of the station. Whether the station can be traced back to the power supply is checked.

[0136] The main function of the power supply tracking is to automatically track the power supply point of the equipment in the station or the feeder equipment outside the station through the network topology tracking function, and to display the power supply path and power supply point on the power grid connection diagram. The power supply point is defined as the equivalent power supply of the substation. The equipment in the station can be a transformer, bus, switch, load, capacitor, etc. The equipment outside the station can be a feeder section, transformer, bus, switch, load, capacitor, etc. The power supply tracking starts from one end of the tracked equipment and performs a depth-first search. When the power supply is found, the search stops. If the power supply is found, the power supply tracking ends, and the power supply and power supply path are colored and displayed on the network. Otherwise, a depth-first search is performed from the other side of the tracked equipment. When the power supply is found, the search stops. If the power supply is found, the power supply tracking ends, and the power supply and power supply path are colored and displayed on the network. Otherwise, the power supply tracking fails. When the tracked equipment is a bus, load, or capacitor, it is a single-ended device, so only one side needs to be searched. When the tracked equipment is a switch or feeder section, it is a double-ended device, so both sides may need to be searched. When the tracked equipment is a transformer, it is supplied from the high-voltage side, so only the high-voltage side needs to be searched.

[0137] The impact user evaluation is directly extracted from all outage distribution transformers in the outage equipment analysis, and the corresponding relationship can be obtained through the mapping in the model. Through cumulative statistics, the corresponding relationship can be obtained.

[0138] The event level is divided into multiple levels from high to low according to the severity of the event consequences.

[0139] The risk rating should analyze the harm of the risk and the possibility of the risk occurrence, comprehensively evaluate the risk size, and determine the risk rating.

[0140] It is worth noting that in the S6, the dynamic prevention and control strategy is generated, a genetic algorithm is used for multi-objective optimization of the maintenance plan, a risk degradation scheme and a prevention and control measure are generated, and the specific steps are as follows:

[0141] D1, the maintenance plan is numbered, the information is fixed in the given time table according to time slicing, a chromosome is created, the time slices are randomly allocated, the plan maintenance single chromosome is generated and encoded, and symbol encoding is adopted;

[0142] D2, the fitness function is established, the comprehensive balance of the maintenance plan is a multi-constraint maintenance optimization problem, while ensuring the safe and economic operation of the power grid, the maximum maintenance willingness of the maintenance unit is met, so as to judge the good and bad of the comprehensive balance of the maintenance plan, the plan maintenance single information and the time slice are input, and the fitness of the maintenance plan is returned after inspection and optimization;

[0143] D3, selection, crossover and mutation operations are performed to generate a new population, strong individuals are selected from the previous generation population to re-encode and generate a new population, selection operation avoids the loss of effective genes, so that high-performance individuals can survive with a larger probability, thereby improving global convergence and computational efficiency, crossover is the most important genetic operation, which simultaneously operates on two chromosomes, combines the characteristics of the two to produce new offspring, and mutation operation refers to replacing the gene value at a certain locus in the individual chromosome code string with other alleles at the same locus, changing the chromosome structure and physical properties, and forming a new individual;

[0144] D4, algorithm termination judgment, mainly including three methods, firstly, the number of cycles, taking the operation time and the performance of the computer as the limiting criterion, generally adopting the method of evolving to the maximum number of generations, secondly, the optimal solution of the objective function, taking the convergence degree of the population in the operation to judge, for example, if the difference between the optimal solution and a solution in the population is controlled within a certain range, it indicates that the evolution is ended, and finally, the change of fitness, taking the comparison of the off-line and on-line performance of the algorithm to judge, and the population fitness does not change much after two or more iterations, which can be stopped.

[0145] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above examples, the above examples and descriptions in the specification are only preferred examples of the present application, and are not intended to limit the present application, various changes and improvements can be made without departing from the spirit and scope of the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent assessment and dynamic prevention and control strategy generation of power grid maintenance risks, characterized in that, Includes the following steps: S1. Construct a knowledge graph-based CIM model of the power grid, obtain the CIM model files of the main grid and distribution network, parse the CIM model files to generate network topology relationships, construct the model with equipment, lines, and electrical equipment containers as graph vertices and relationships as edges connecting vertices, and store it in a graph database. S2. Multi-source data fusion and feature extraction to build a feature model. Collect maintenance plans, real-time operation data, equipment defect database and historical risk data. Use LSTM multivariate time series with multiple variables to analyze and predict load characteristics, build a load characteristic feature model, consider important users and large users to build a load distribution feature model, and build a power supply capacity feature model, maintenance capacity feature model and maintenance capacity feature model. S3. Power grid topology analysis: Depth-first search is used to find the main lines and loops in the power grid, determine the hierarchical relationship and connection method between the lines, and breadth-first search is used to find the branches and terminal nodes in the power grid, establish the branch structure and hierarchical relationship of the power grid, and perform search for transferable loads, N-1 criterion analysis and power failure analysis. S4. Verification of maintenance constraint rules: The verification of maintenance constraint rules includes safety constraints, coordination constraints, maintenance resource constraints, equipment constraints, and economic constraints. Relationships are established using knowledge graph technology. S5. Intelligent risk assessment: Based on the power grid CIM model and power grid topology analysis, combined with equipment health status and weather influencing factors, it assesses the maintenance risk level and identifies the health status of key equipment. S6. Dynamic prevention and control strategy generation: Genetic algorithm is used to optimize the maintenance plan for multiple objectives and generate risk reduction schemes and prevention and control measures.

2. The method for intelligent assessment and dynamic prevention and control strategy generation of power grid maintenance risks according to claim 1, characterized in that: In step S1, when constructing the knowledge graph model, pruning optimization is performed, endpoints are pruned, and connection points are directly accessed through devices to shorten the topology path and reduce the number of graph nodes.

3. The method for intelligent assessment and dynamic prevention and control strategy generation of power grid maintenance risks according to claim 1, characterized in that, In step S2, a multivariate LSTM time series model incorporating multiple variables is used for load characteristic analysis and prediction. The specific steps are as follows: A1. Collect historical power load data, divide it into multiple time steps according to the time series, and the input data format is a time series vector. The shape of each time step's data is (batch_size, input_dim), where batch_size is the batch size and input_dim is the feature dimension. A2. Remove abnormal load data and normalize the data; A3. Construction of the core structure of the LSTM model, with input including cell state C. t-1 Hidden state h t-1 and the input vector x at time t t Output the cell state C at the next time step. t and hidden state h t Both cell states and hidden layer states are initialized as all-zero vectors. Each neural network layer has three gate structures—forget gate, input gate, and output gate—to protect and control information. The forget gate filters historical information, the input gate generates candidate states and updates the cell state, and the output gate determines the current output. The corresponding formula is: f t =σ(w f [h t-1 ,x t ]+b f ) i t =σ(w i [h t-1 ,x t ]+b i ) the t =σ(w o [h t-1 ,x t ]+b o ) h t = no t fishy(C) t ) In the formula, f t The output of the forget gate is represented by σ, which represents the sigmoid function, and w is the output of the forget gate. f b represents the forget gate weight. f Indicates the forget gate bias, [h t-1 ,x t ] indicates concatenating two vectors, i t w represents the output of the input gate. i Indicates the input gate weight, b i This indicates the input gate bias. Represents the candidate cell state, tanh represents the activation function, and w c The calculated weights representing the candidate cell states, b c The computational bias representing the candidate cell state, o t Indicates the output of the output gate, w o Indicates the output gate weight, b o This indicates the output gate bias.

4. The method for intelligent assessment and dynamic prevention and control strategy generation of power grid maintenance risks according to claim 3, characterized in that, In S2, a power supply capacity characteristic model is constructed based on the maximum power supply capacity. According to the capacity of the main transformer and feeders in the power grid and the network topology, the maximum load supply capacity of the feeders is calculated, and a comprehensive evaluation is made from the aspects of the main transformer power supply capacity, feeder comprehensive current, line bottleneck current, CT ratio, and the highest current in the past three months. In the formula, S max S represents the maximum power of the feeder. 3,max This indicates the maximum power of the feeder over the past three months, S x Indicates the power that needs to be transferred, I max I represents the maximum supply current of the feeder. 3,max I represents the highest current in the feeder over the past three months. x Indicates the current that needs to be supplied, S B Indicates the main transformer's power supply capacity, I lc This indicates the total current carrying capacity of the feeder. Indicates the maximum allowable current carrying capacity of the wire diameter, I ct This represents the CT ratio of the feeder, and min(·) indicates taking the minimum value; Main transformer power supply capacity S B The calculation formula depends on the capacity of the main transformer within the substation: In the formula, S in This indicates the rated capacity of the transformer's high-voltage side, S. out This indicates the rated capacity of the low-voltage side of the transformer. Indicates the power factor; The CT ratio is the ratio between the converted currents on both sides of the current transformer.

5. The method for intelligent assessment and dynamic prevention and control strategy generation of power grid maintenance risks according to claim 4, characterized in that, In S2, the specific steps for constructing the maintenance capability characteristic model are as follows: B1. Collect the number of out-of-power equipment, the number of safety measure groups, the number of distribution automation switches, and the operation time of each department from the power grid maintenance records to form a structured dataset; B2. Clean the data, standardize the time unit and equipment type classification, and remove outliers; B3. Determine the input and output variables. Input variable X includes the number of outage equipment X1, the number of safety measure groups X2, and the number of distribution automation switches X3. Output variable Y includes the installation time of safety measures in the station Y1, the removal time of safety measures in the station Y2, the power outage operation time Y3, the power restoration operation time Y4, the power outage order time Y5, and the power restoration order time Y6. B4. Establish a linear regression model and adopt the multiple linear regression equation Y=β0+β1X1+β2X2+β3X3+ε, where β0 is the intercept, β1, β2 and β3 are all regression coefficients, and ε is the error term. B5. Use the least squares method to solve for the coefficients, and fit the parameters using historical data to minimize the sum of squared errors between the predicted and actual values.

6. The method for intelligent assessment and dynamic prevention and control strategy generation of power grid maintenance risks according to claim 5, characterized in that, In S2, a maintenance capacity characteristic model is constructed. The ratio of the reserved reserve capacity to the maximum operating capacity of the online operating power stations participating in power generation dispatch is equal to the ratio of the total system reserve capacity to the annual maximum load of the power system. The sum of the reserved reserve capacities of all power stations equals the total reserve capacity of the power system, i.e.: In the formula, S A S represents the reserve capacity reserved at the power station, and S represents the maximum operating capacity of the system. D S represents the total system reserve capacity. M This indicates the system's maximum annual load.

7. The method for intelligent assessment and dynamic prevention and control strategy generation of power grid maintenance risks according to claim 1, characterized in that: In step S3, available load search, N-1 criterion analysis, and power failure analysis are performed. Available load search involves graph traversal calculation of power failure equipment using a depth-first search algorithm to trace available power sources. N-1 criterion analysis uses the minimum spanning tree algorithm to analyze and traverse connection relationships. Power failure analysis uses the minimum spanning tree algorithm to traverse the set of equipment that has failed after the equipment is currently in operation, and then performs complement calculation on the two sets to obtain the current set of power failure equipment.

8. The method for intelligent assessment and dynamic prevention and control strategy generation of power grid maintenance risks according to claim 1, characterized in that: In S4, the verification of maintenance constraint rules includes safety constraints, coordination constraints, maintenance resource constraints, equipment constraints, and economic constraints. Safety constraints include active power flow constraints, transfer capacity constraints, and protection action verification. Coordination constraints include simultaneous maintenance constraints, sequential maintenance constraints, mutually exclusive maintenance constraints, and non-changeable maintenance constraints. Maintenance resource constraints include maintenance unit constraints and dispatch resource constraints. Equipment constraints include transmission and transformation combination constraints and the maximum number of maintenance tie lines. Economic constraints include reducing load loss and reducing power outage duration.

9. The method for intelligent assessment and dynamic prevention and control strategy generation of power grid maintenance risks according to claim 1, characterized in that, In step S5, based on the power grid CIM model and power grid topology analysis, combined with equipment health status and weather influencing factors, the maintenance risk level is assessed, and the health status of key equipment is identified. The specific steps are as follows: C1. Calculate the real-time failure rate of the equipment based on the health index of the equipment status; C2. The following steps are performed in sequence: assess the reduced load, identify the substations experiencing power loss, assess the impact on users, and determine the event level and risk level.

10. The method for intelligent assessment and dynamic prevention and control strategy generation of power grid maintenance risks according to claim 1, characterized in that, In step S6, the dynamic prevention and control strategy is generated by using a genetic algorithm to perform multi-objective optimization of the maintenance plan, generating risk degradation schemes and prevention and control measures. The specific steps are as follows: D1. Number the maintenance plan, divide it into time slices, fix the information in the given time schedule, create a chromosome, randomly allocate time slices, generate a single chromosome for planned maintenance and encode it. D2. Establish a fitness function, input the planned maintenance order information and time segment, and after verification and optimization, return the fitness of the maintenance plan; D3. Perform selection, crossover, and mutation operations to generate a new population; D4. Algorithm termination judgment.

Images