New power system restoration decision method and system considering the influence of unascertained faults

Abstract

The application belongs to the field of new power system restoration, and provides a new power system restoration decision method and system considering uncertain fault influence, so as to take generator output, load restoration quantity and load loss as objective functions, construct a new power system restoration optimization model, screen line fault state certain scenes higher than a predetermined probability according to line fault probability distribution, respectively solve optimal restoration decisions of each scene, and establish a pre-optimization decision set, construct a decision index in two dimensions of efficiency and risk, construct a Monte Carlo tree search model guided by pre-optimization, preferentially take operations matched with the current state in the pre-optimization decision set as expansion candidates, take the decision index as a reward value, and update the access frequency and cumulative value of the search tree node through back propagation, and search and roll output the restoration decision online. The application can cope with uncertain conditions of line fault state.

Description

Technical Field

[0001] This invention belongs to the field of novel power system restoration, specifically relating to a novel power system restoration decision-making method and system that takes into account the impact of unknown faults. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] The ability to restore the power grid structure and critical loads in the shortest possible time and in a reasonable order after a major power outage directly affects the duration of the power outage and the scope of the accident's impact. Therefore, rapid and reliable restoration decisions have become an indispensable key capability for the dispatch and operation of new power systems.

[0004] However, in actual large-scale power outage scenarios, factors such as the spread of the accident, the paralysis of information systems, and the interruption of communication links often make it difficult for dispatchers to grasp the real-time status of each line in a timely and accurate manner, and it is unknown whether some affected lines can still be put into operation. If lines with uncertain status are conservatively regarded as faulty lines, the risk of accidental commissioning can be reduced, but it may lead to insufficient utilization of network reconstruction resources and delay the recovery process. If they are regarded as intact lines, the failure of power-on test will waste human and material resources. Even worse, accidental commissioning of faulty lines may cause short circuits or line breaks, leading to the failure of low voltage ride-through (LVRT) of restored wind farms and disconnection from the grid, further increasing the risk of system operation and economic losses. Therefore, the uncertainty of the status of transmission lines after a disaster cannot be ignored when making decisions on the recovery of new power systems.

[0005] Currently, the recovery decision-making technologies used in engineering are mostly geared towards scenarios where the status of the lines is known. The dispatching department mainly carries out recovery operations based on pre-established black start plans and dispatching procedures, and offline arranges a deterministic recovery sequence for dispatchers to call on-site. However, in actual large-scale power outage scenarios, factors such as communication interruption and information system paralysis often prevent the real-time status of some affected lines from being transmitted in a timely manner, rendering the premise of known status of the plan invalid. Dispatchers can only make short-term judgments based on experience, which carries the risk of mistakenly putting faulty lines into operation, causing wind farms to disconnect from the grid and loads to be cut off again. It is also difficult to quickly weigh the recovery efficiency against potential risks in multiple operations. Summary of the Invention

[0006] To address the aforementioned problems, this invention proposes a novel power system restoration decision-making method and system that takes into account the impact of unknown faults. This invention can generate optimal restoration decisions online that balance safety and restoration benefits under conditions of uncertain line fault status.

[0007] According to some embodiments, the present invention adopts the following technical solution: A novel power system restoration decision-making method that takes into account the impact of unknown faults includes the following steps: A novel power system recovery optimization model is constructed, taking generator output, load recovery amount, and load loss amount as objective functions, and considering the impact of wind farm disconnection and line breakage faults. Based on the line fault probability distribution, known scenarios of line fault states with a probability higher than the predetermined probability are selected. Based on the novel power system recovery optimization model, the optimal recovery decision for each scenario is solved, and a pre-optimized decision set is established. We construct decision indicators with two dimensions: efficiency and risk, for scenarios where the state of line faults is unknown. The efficiency dimension focuses on the efficiency of load recovery, while the risk dimension focuses on the negative impacts of wind farm disconnection from the grid and line outages. A pre-optimized guided Monte Carlo tree search model is constructed. In the selection and expansion phase, operations that match the current state in the pre-optimized decision set are given priority as expansion candidates. In the simulation phase, a directed in-depth simulation strategy that considers the probability of line failure is formulated. In the backtracking phase, the decision index is used as the reward value, and the number of visits and cumulative value of the search tree nodes are updated through backpropagation. The Monte Carlo tree search model is used to search online and output recovery decisions in a rolling manner. The recovery operation corresponding to the child node with the most visits under the root node is selected as the optimal recovery action for the current time step. After executing the optimal recovery action, the state of the power system is updated. The subsequent recovery decision is determined in a rolling manner in the scenario where the line state is unknown.

[0008] As an alternative implementation method, the process of constructing a new power system recovery optimization model that considers the impact of wind farm disconnection and line breakage faults, with generator output, load recovery amount, and load loss amount as objective functions, includes: optimizing the recovery process from three perspectives: generator output, load recovery, and load loss, with the objective function being the optimal weighted sum of each scenario, and setting connectivity constraints, component recovery sequence constraints, conventional unit output constraints, wind turbine output constraints, and node power balance constraints, and considering the impact of wind farm disconnection and line breakage faults in the process of setting the constraints.

[0009] As an alternative implementation method, the process of screening known scenarios of line fault states with a probability higher than a predetermined probability based on the line fault probability distribution includes: calculating the joint probability of each scenario and selecting the scenarios with the highest joint probability. K The scenario constitutes a known scenario of a line fault state with a higher probability than expected. ， parameter K The value is determined based on the scale of the power system and computing resources; The joint probability is calculated based on the fault probability and the probability of each line being in a normal state.

[0010] As an alternative implementation method, the process of solving the optimal recovery decision for each scenario and establishing a pre-optimized decision set includes: for each scenario in the scenario where the line fault state is known to be higher than the predetermined probability, based on the established new power system recovery optimization model, using the mixed integer programming method to solve its optimal recovery decision in the current recovery period, and collecting the optimal recovery decisions of each scenario in each period to establish a pre-optimized decision set as a guiding prior.

[0011] As an alternative implementation method, the process of constructing decision indicators in the dual dimensions of efficiency and risk for scenarios where the line fault state is unknown includes: the decision indicators are: ; Among them, the risk indicators of load shedding caused by wind farm disconnection from the grid are: ; ; Load-benefit metrics for recovery decisions: ; In the formula, α , β These are the weighting coefficients for load revenue and grid disconnection risk, respectively. For the current decision-making step; i index for bus node N For the set of bus nodes; W For the collection of wind farms in the system; t Index for system recovery period; The future time step length in the MCTS simulation; The representative is the first s Under the simulated trajectory t Periodic wind farm w Whether it is offline or not Offline Not offline; indicator variable The value is 1 when at least one wind farm goes offline; For the first s Under the simulated trajectory t Time period nodes The amount of load that has been restored; For the first s The cumulative amount of unloaded load simulated within the future time step under the simulated trajectory; For the first s The simulated load recovery amount within the future time step under the simulated trajectory The summation of variables at each time step effectively introduces a time-weighted effect; loads restored in earlier time periods continue to supply power in subsequent time periods, contributing more to the cumulative index. It is regarded as a comprehensive performance indicator for measuring the power restoration speed and timeliness of the restoration plan.

[0012] As an alternative implementation, the process of constructing a pre-optimized guided Monte Carlo tree search model includes: the Monte Carlo tree search model gradually constructs a search tree by repeatedly executing four stages: selection, expansion, simulation, and backtracking, and finally selects the action with the most visits as the optimal decision. In the selection stage, based on the established pre-optimized decision set, with the current recovery state as the root node and candidate recovery lines as branches, when the number of iterations is within a predetermined range, the search process is in the prior guidance stage, and actions in the optimal recovery sequence of fault scenarios with a higher probability than the predetermined probability are selected as priority candidates; when the number of iterations exceeds the predetermined range, the search process enters the global search stage, and selection is performed according to topological connectivity. During the expansion phase, from the lines that are not yet in operation in the current state, lines that meet the connectivity constraints are selected as the candidate action set, set as the child nodes of the current node, and their upper confidence interval index value is set to infinity. The new node inherits the state information of the parent node and updates the set of lines that are already in operation.

[0013] As a further defined implementation method, the process of constructing a pre-optimized guided Monte Carlo tree search model includes: during the simulation phase, a directed in-depth simulation strategy that takes into account the probability sampling of line faults is adopted, and the model continues to expand along the frontier direction of the currently restored topology in order to enhance the topological continuity of the simulation path; if there are no more lines to extend in the current direction, the model then returns to the global set of available actions for selection. During the backtracking phase, the reward values ​​obtained in the simulation phase are propagated backward along the search path to update the statistical information of each node on the path. The simulation is iterated until the accumulated statistical information can reflect the expected benefits and risks of different recovery paths under failure uncertainty.

[0014] As an alternative implementation, the process of using the Monte Carlo tree search model to search online and output recovery decisions includes: embedding a wind farm disconnection fast judgment device in the Monte Carlo tree search model, which is used to judge the wind farm disconnection status under the candidate recovery operation or recovery path by taking the network topology information, fault location and fault duration after the candidate recovery operation or recovery path is executed as input.

[0015] As an optional implementation, the process of using the Monte Carlo tree search model to search online and output recovery decisions in a rolling manner includes: taking the candidate recovery operations formed in the simulation phase and the off-grid status of each wind farm as input, substituting them into the new power system recovery optimization model, calculating the decision indicators corresponding to the candidate recovery operations or recovery paths, and using them as the reward value of this simulation to propagate back along the search path, updating the access count and cumulative value of each node in the search tree; after reaching the predetermined number of search iterations or the preset search time limit, selecting the recovery operation corresponding to the child node with the most access count under the root node of the search tree as the optimal recovery action for the current time step, executing the optimal recovery action and synchronously updating the power system status; using the updated system status as the initial state of the root node of the new round of search, performing cyclic execution, and outputting subsequent recovery decisions in scenarios where the line status is unknown in a rolling manner; The entire recovery decision-making process after a major power outage is completed until all components to be restored are put into operation or the recovery termination conditions stipulated by the dispatching authority are met.

[0016] A novel power system restoration decision-making system that takes into account the impact of unknown faults includes: The system recovery optimization model construction module is configured to use generator output, load recovery amount and load loss amount as objective functions to construct a new power system recovery optimization model that considers the impact of wind farm disconnection and line failure. The pre-optimized decision set solving module is configured to filter known scenarios of line fault states with a probability higher than a predetermined probability based on the line fault probability distribution, and solve the optimal recovery decision for each scenario based on the new power system recovery optimization model to establish a pre-optimized decision set. The decision indicator construction module is configured to construct decision indicators in two dimensions: efficiency and risk, under the scenario where the line fault status is unknown. The efficiency dimension focuses on the efficiency of load recovery, while the risk dimension focuses on the negative impact of wind farm disconnection from the grid and line outage. The pre-optimized guided Monte Carlo tree search module is configured to build a pre-optimized guided Monte Carlo tree search model. In the selection and expansion phase, operations that match the current state in the pre-optimized decision set are given priority as expansion candidates. In the simulation phase, a directed in-depth simulation strategy that considers the probability of line failure is formulated. In the backtracking phase, the decision index is used as the reward value, and the number of visits and cumulative value of the search tree nodes are updated through backpropagation. The online search and rolling decision output module is configured to use the Monte Carlo tree search model to search online and output recovery decisions in a rolling manner. It selects the recovery operation corresponding to the child node with the most visits under the root node as the optimal recovery action for the current time step, and updates the state of the power system after executing the optimal recovery action. It determines the subsequent recovery decision in the scenario where the line state is unknown in a rolling manner.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention innovatively constructs a comprehensive evaluation index with two dimensions: efficiency and risk. It also takes into account the cumulative load recovery efficiency brought about by the successful commissioning of the line, as well as the load loss risk caused by the wind farm disconnecting from the grid due to the faulty line being commissioned. These factors are used as the reward value in the Monte Carlo tree search backtracking stage to participate in the node value update, and the output is a recovery decision that can weigh recovery efficiency and risk under the condition that the fault state is unknown.

[0018] The pre-optimized guided Monte Carlo tree search model proposed in this invention obtains the prior recovery operation sequence for each scenario through a recovery decision optimization model that takes into account the effects of short circuits and disconnections. The operations in the prior sequence are used as priority candidates and their feasibility is verified. In the simulation stage, a prior-guided simulation strategy is adopted, which effectively avoids the problem of randomly selecting a large number of unreasonable recovery sequences in the early stage of the search, and significantly improves the convergence speed and decision quality of online search.

[0019] This invention proposes a simulation strategy that considers the probability of line faults. The fault state of the currently selected candidate line is sampled and a random number is generated. If the random number is less than a set value, the sampled state is a fault state; otherwise, the sampled state is a normal state. This method can effectively handle the decision uncertainty caused by the unknown state of the line and avoid the dimensionality curse caused by the traditional deterministic optimization method of directly enumerating all fault combinations.

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0021] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0022] Figure 1 A flowchart illustrating the overall process of a novel power system recovery decision-making method that takes into account the impact of unknown faults, according to one embodiment; Figure 2 is a flowchart of a pre-optimized guided Monte Carlo tree search framework according to one embodiment. Detailed Implementation

[0023] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0024] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0025] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0026] Where there is no conflict, the embodiments and features described in this application may be combined with each other.

[0027] Example 1 A novel power system recovery decision-making method that takes into account the impact of unknown faults, such as Figure 1 As shown, it includes the following steps: S1. Construct a system recovery optimization model that considers the impact of wind farm grid disconnection and line failure. The objective function of the model comprehensively considers generator output, load recovery amount and load loss amount. The constraints include connectivity constraints, component recovery sequence constraints, conventional unit output constraints, wind turbine output constraints and node power balance constraints. S2. Establish a pre-optimized decision set. Based on the line fault probability distribution, select high-probability known line fault states and solve the optimal recovery decision for each scenario to establish the pre-optimized decision set.

[0028] S3. Propose an "efficiency-risk" decision indicator for scenarios where the line fault status is unknown. The efficiency dimension focuses on the efficiency of load recovery, while the risk dimension focuses on the negative impact of wind farm disconnection and line breakage.

[0029] S4. Construct a pre-optimization guided Monte Carlo Tree Search (MCTS) framework. In the selection-expansion phase, prioritize operations in the pre-optimization decision set output in step S2 that match the current state as expansion candidates. In the simulation phase, formulate a directed in-depth simulation strategy that considers the probability of line failure. In the backtracking phase, use the decision indicators proposed in step S3 as reward values ​​and update the number of visits and cumulative value of the search tree nodes through backpropagation.

[0030] S5. Online search and rolling output of recovery decisions: A fast wind farm disconnection judgment device is embedded within a pre-optimized Monte Carlo tree search framework to quickly determine the wind farm disconnection status under candidate recovery operations or recovery paths. Simulated candidate recovery operations are used as known inputs into the system recovery optimization model to calculate the corresponding efficiency-risk decision indicators. The recovery operation corresponding to the child node with the most visits under the root node is selected as the optimal recovery action for the current time step. After executing this recovery action, the system state is updated, and subsequent recovery decisions for scenarios where the line status is uncertain are output in a rolling manner.

[0031] In step S1, a system recovery optimization model considering the impact of wind farm grid disconnection and line breakage faults is constructed. The objective function and constraints of the model are as follows: Objective function: For grid restoration under uncertain line faults, the restoration process needs to be optimized by considering three aspects: generator output, load restoration, and load loss, and by using the weighted sum of these factors across all scenarios. No. s Maximizing generator output under the simulated trajectory: (1) No. s Maximizing load recovery under the simulated trajectory: (2) No. s Minimize load loss under the simulated trajectory: (3) The overall objective function is: (4) In the formula: For the first s Unit under the simulated trajectory g exist t The amount of effort contributed during a given period; For the unit exist t The start-up status variable for a time period is set to 1 if the unit starts during that time period, and 0 otherwise. For the first s Nodes under the simulated trajectory i Load at t Recovery amount over a period of time; For the first s Nodes under the simulated trajectory i Load at t The amount of load loss during the time period; , , These are the weighting coefficients for generator output, load recovery, and load loss, respectively. This refers to the unit start-up bonus coefficient. The future time step length is the simulation length of MCTS. , These are the weighting coefficients of unit output revenue, load recovery revenue, and load loss risk in the comprehensive objective function, respectively. , , These represent the three sub-objective functions: unit output revenue, load recovery revenue, and load loss risk, respectively.

[0032] Connectivity constraints: In recovery decisions, connectivity constraints are mainly used to restrict whether a node is recovered. They are key to ensuring that recovery actions (such as putting lines into operation and starting generators) are feasible and maintaining network topology connectivity.

[0033] (5) (6) (7) (8) (9) (10) (11) in N It is the set of bus nodes; N B This refers to the bus node where the black start power supply is located; L S is the set of system circuits; S is the set of decisions made during the simulation phase. i, j It is the index of the bus node; t It is the index for the system recovery period; a Boolean variable. Representing the s Under the simulated trajectory t Time-of-day routes ij The decision plan, 1 indicates that the route is planned to be restored, and 0 indicates that the route is not planned to be restored; Boolean variable. Representing the s The route under the simulated trajectory ij The sampling fault status is 1, which represents that the line sampling is a normal line, and 0 represents that the line sampling is a short circuit + open line fault line. For the first s Under the simulated trajectory t The virtual commodity flow variable in the time-segment planning layer is used to characterize the connectivity relationship between nodes and the black-start root node, but does not represent the actual power flow. n bus = | N| represents the total number of buses, used as an upper bound constant for the Big-M type; It represents the maximum number of lines put into operation within a single time period.

[0034] Component recovery sequence constraints: (12) (13) (14) (15) (16) in G A collection of conventional generating units; G B This is a collection of black start generator sets; For the first s The simulated trajectory is below. t Time-of-use units g Whether to enter the startup phase (0-1 variable) start up, Not started); For the first s Under the simulated trajectory t Periodic wind farm w The 0-1 variable indicating whether it has entered the grid-connected operation phase. Grid connection, (Not connected to the grid).

[0035] Conventional unit output model Using an ideal ramp-up curve model, the unit startup process is divided into a startup preparation phase and a grid connection operation phase: (17) (18) (19) (20) (twenty one) in T g Number of steps during unit startup; Boolean variable Representing the s Under the simulated trajectory t Time-of-use units g Whether it is connected to the grid; Representing the s Unit at time t under the simulated trajectory g Those who have made contributions Representing the s The unit in the previous time period under the next simulated trajectoryg Those who have made contributions This represents the gradeability of a conventional generator set. This represents the downhill slope rate of a conventional generator set. M This is a constant value, with a value of 100. m This is an auxiliary constant with a value of 0.995.

[0036] Wind turbine output model wind farm w In the s Under the simulated trajectory Actual output during the period satisfy: (twenty two) (twenty three) in, Indicates time period wind farm The maximum predicted available output is determined by the wind speed forecast and the wind turbine power curve. W A collection of wind turbine units; Representing the s Under the simulated trajectory t Periodic wind farm w Is it offline? Offline (Not offline) The representative is the first s Under the simulated trajectory t Periodic wind farm w Maximum predicted output; T r This represents the minimum number of steps required to restore lock-up after a wind farm is disconnected from the grid.

[0037] Node power balance constraints: (twenty four) (25) (26) (27) (28) (29) in, For the first s The route under the simulated trajectory ij The positive trend on the frontier; For the first s The route under the simulated trajectory ij The unproductive current on the surface; For the firsts Nodes under the simulated trajectory i , j The voltage phase angle difference between them; , The first s Under the simulated trajectory t Time-of-use units g The effective and ineffective contributions; Representing the s Under the simulated trajectory t Time period nodes i The voltage amplitude; , The first s Under the simulated trajectory t Time-of-use wind turbine w The effective and ineffective contributions; W A collection of wind turbine units; For the first s Under the simulated trajectory t Time period nodes The amount of load that has been restored; For nodes i The unit's starting power; tanφ i For nodes i The power factor of the load. This represents the maximum permissible value for the phase angle difference.

[0038] The above constraints define the active / reactive power flow of the lines (only lines that have been restored and are fault-free are allowed to flow), phase angle difference safety constraints, and node constraints, respectively. i The range of active / reactive power balance and load recovery.

[0039] In step S2, during the power system recovery process after a major power outage, each line to be restored may be in a normal or faulty state. Considering that the number of fault scenarios increases exponentially with the number of lines, to reduce computational complexity, this invention employs the following strategy to screen high-probability scenarios, calculate the joint probability of each scenario, and select the scenarios with the highest joint probabilities. K The scenarios constitute a high-probability failure scenario set S K .parameter K The value of can be determined based on the system size and computing resources, with a typical range of 50-200.

[0040] Assume the system has a total of N One line to be restored i - j The probability of line failure is p i-j The probability that the line is in a normal state is 1- pi-j .

[0041] Because line faults are independent, any combination of fault scenarios k The joint probability can be calculated as follows: (30) in, F k For the scene k The set of lines in a faulty state. For each scenario in the set. k Based on the system recovery optimization model established in step S1, a mixed integer programming method is used to solve its solution during the current recovery period. t Optimal recovery decision under The optimal recovery decisions for each scenario at each time period are aggregated to establish a pre-optimized decision set, which serves as a guiding prior for the early extension of the PMCTS framework in step S4.

[0042] In step S3, for scenarios where the line fault status is unknown, the decision-making effectiveness during the recovery process can be evaluated from two dimensions: benefit and risk. The benefit dimension focuses on the efficiency of load recovery, while the risk dimension focuses on the negative impacts of wind farm disconnection and line outages. The efficiency-risk decision-making index is defined as follows: (31) Among them, the risk indicators of load shedding caused by wind farm disconnection from the grid are: (32) (33) Load-benefit metrics for recovery decisions: (34) In the formula: α , β These are the weighting coefficients for load revenue and grid disconnection risk, respectively. The current decision step; indicator variable The value is 1 when at least one wind farm goes offline; For the first s The cumulative amount of unloaded load simulated within the future time step under the simulated trajectory; No. s The summation of load restoration within the simulated future time steps under the simulated trajectory; the summation of variables at each time step effectively introduces a time-weighted effect, meaning that load restored in earlier time periods continues to supply power in subsequent time periods, contributing more to the cumulative index. It can be regarded as a comprehensive performance indicator for measuring the power restoration speed and timeliness of the restoration plan.

[0043] In step S4, a pre-optimized guided Monte Carlo tree search framework is constructed to search online for recovery operations under the current actual system state. Monte Carlo tree search is a sampling-based online planning algorithm that gradually builds a search tree by repeatedly executing four stages: "selection—expansion—simulation—backtracking," ultimately selecting the action with the most visits as the optimal decision. Figure 2 As shown, the search process in this embodiment is as follows: (1) Selection stage Using the current actual system state as the root node and the available recovery routes as branches, a Monte Carlo tree search framework guided by pre-optimization is constructed based on the pre-optimized decision set established by S2. This framework uses the current recovery state as the root node and candidate recovery routes as branches, with the number of iterations... n satisfy n At this point, the search process is in the prior guidance phase, prioritizing actions in the optimal recovery sequence for high-probability fault scenarios; when the number of iterations... n Satisfying n At this point, the search process enters the global search phase, where the selection is made according to the standard UCT formula and topological connectivity.

[0044] Mixed UCT formulas including prior guiding terms: (35) in: (36) Prior guiding function Based on the joint probability of relevant failure scenarios Each recommendation operation is weighted. Decay weights are applied. Adaptive linear decay form is adopted: (37) in, and Each represents an action Cumulative rewards and number of visits; Indicates the number of times the parent node has been visited; Indicates action The number of times it was accessed or selected; For the exploration coefficient; For the first Current system state at the current time step; For state The set of candidate operations; Used for weighted prior guiding terms; This is an indicator function; when the condition inside the parentheses is true, ;otherwise, . Indicating a fault scenario Below, the first obtained through the pre-optimization model Timewalk recommends recovery exercises; Representing a scene The probability of their joint occurrence; This is the initial attenuation coefficient; This represents the number of iterations. and These represent the number of fault scenarios and the number of pre-search iterations for each scenario, respectively. (2) Expansion phase From the lines that are not yet in operation in the current state, select the lines that meet the connectivity constraints as the candidate action set, set them as child nodes of the current node, and set their upper confidence interval index value to infinity. The new node inherits the state information of the parent node and updates the set of lines that are in operation.

[0045] (3) Simulation phase To improve the simulation's ability to depict the power system recovery process, this paper adopts a directed depth simulation strategy that considers line fault probability sampling during the rollout phase. This strategy prioritizes expanding along the leading edge of the currently restored topology to enhance the topological continuity of the simulation path; if no further lines can be extended in the current direction, it then retreats to the global set of available actions for selection. This strategy better reflects the actual characteristic of power system recovery: "gradually expanding outward along the energized area."

[0046] Set time step The system status is The energized busbars are assembled as follows The system's set of actions is If time step The recovery action performed was for the line. The set of leading-edge buses newly restored by this action is defined as: (38) Based on the set of leading-edge buses, the set of candidate actions for continuing to expand along the current leading-edge direction is defined as follows: (39) In the formula, in the formula, Indicates time step The set of leading-edge buses, that is, new energized nodes that are directly energized by the current recovery action and can serve as the starting point for subsequent topology expansion; Indicates time step The collection of energized busbars; Indicates time step The selected recovery route. Representing state The set of candidate actions at the forefront; Represents the set of globally available actions; This indicates that the starting point of the candidate route is located in the current set of leading-edge buses; This indicates that the candidate line's endpoint is a non-energized bus that has not yet been restored.

[0047] Directed Deepening Candidate Selection: Based on the frontier bus set of Equation (38) and the frontier candidate action set of Equation (39), the time step is selected according to the following probability rule. t +1 simulated action: (40) In the formula, Indicates time step Simulated actions; Indicates the state Select action The conditional probability; and These represent the size of the frontier candidate action set and the global action set, respectively.

[0048] Fault probability sampling: Assume simulated action Corresponding candidate routes ij The failure probability is p i-j For the currently selected candidate route ij Perform fault state sampling and generate random numbers. r If ∈[0,1], r<p i-j If the sampled condition is faulty, it is considered a normal condition; otherwise, it is considered a normal condition.

[0049] Bus set update restored: Selected simulation action After the sampling is confirmed to be in a normal state, the next step is to update the set of recovered buses as follows: (41) In the formula, Indicates the execution of an action The newly formed frontier busbar set; This indicates that the starting point of the line is the currently energized busbar; This indicates that the line terminates at a busbar that was not energized before restoration. Repeat the above to simulate... T w At each time step, the efficiency and risk items of the trajectory are accumulated, and the comprehensive efficiency-risk index is calculated as the return value of this simulation.

[0050] (4) Retrospective phase The reward values ​​obtained in the simulation phase are propagated back along the search path to update the statistics of each node on the path. Since the line fault state is probabilistic, the same line may exhibit different fault states in different simulation trajectories. Therefore, a large number of simulation iterations are required, and the accumulated statistical information can reflect the expected benefits and risks of different recovery paths under fault uncertainty.

[0051] In step S5, a wind farm disconnection fast judgment device is embedded in the PMCTS framework. The network topology information, fault location and fault duration after the candidate recovery operation or recovery path is executed are used as inputs to quickly judge the wind farm disconnection status under the candidate recovery operation or recovery path.

[0052] The candidate recovery operations and the grid disconnection status of each wind farm formed in the simulation phase are used as inputs. These are then substituted into the system recovery optimization model constructed in step S1, which considers the impact of wind farm grid disconnection and line breakage faults. The efficiency-risk comprehensive decision index corresponding to the candidate recovery operation or recovery path is calculated and used as the reward value for this simulation. It is then propagated back along the search path to update the access count and cumulative value of each node in the search tree. After reaching the predetermined number of search iterations or the preset search time limit, the recovery operation corresponding to the child node with the most access count under the root node of the search tree is selected as the optimal recovery action for the current time step. The scheduler executes the recovery action and updates the system status synchronously. The updated system status is used as the initial state of the root node for the new round of search, and the process returns to step S4. This process is repeated cyclically to output subsequent recovery decisions in scenarios where the line status is unknown in a rolling manner.

[0053] The entire recovery decision-making process after a major power outage is completed until all components to be restored are put into operation or the recovery termination conditions stipulated by the dispatching authority are met.

[0054] Example 2 A novel power system restoration decision-making system that takes into account the impact of unknown faults includes: The system recovery optimization model construction module is configured to use generator output, load recovery amount and load loss amount as objective functions to construct a new power system recovery optimization model that considers the impact of wind farm disconnection and line failure. The pre-optimized decision set solving module is configured to filter known scenarios of line fault states with a probability higher than a predetermined probability based on the line fault probability distribution, and solve the optimal recovery decision for each scenario based on the new power system recovery optimization model to establish a pre-optimized decision set. The decision indicator construction module is configured to construct decision indicators in two dimensions: efficiency and risk, under the scenario where the line fault status is unknown. The efficiency dimension focuses on the efficiency of load recovery, while the risk dimension focuses on the negative impact of wind farm disconnection from the grid and line outage. The pre-optimized guided Monte Carlo tree search module is configured to build a pre-optimized guided Monte Carlo tree search model. In the selection and expansion phase, operations that match the current state in the pre-optimized decision set are given priority as expansion candidates. In the simulation phase, a directed in-depth simulation strategy that considers the probability of line failure is formulated. In the backtracking phase, the decision index is used as the reward value, and the number of visits and cumulative value of the search tree nodes are updated through backpropagation. The online search and rolling decision output module is configured to use the Monte Carlo tree search model to search online and output recovery decisions in a rolling manner. It selects the recovery operation corresponding to the child node with the most visits under the root node as the optimal recovery action for the current time step, and updates the state of the power system after executing the optimal recovery action. It determines the subsequent recovery decision in the scenario where the line state is unknown in a rolling manner.

[0055] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of one or more computer-usable storage media (including, but not limited to, disk storage, etc.) containing computer-usable program code. CD - ROM It takes the form of a computer program product implemented on (such as optical memory, etc.).

[0056] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0057] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0058] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0059] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made by those skilled in the art without creative effort within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A novel power system restoration decision-making method that takes into account the impact of unknown faults, characterized in that, Includes the following steps: A novel power system recovery optimization model is constructed, taking generator output, load recovery amount, and load loss amount as objective functions, and considering the impact of wind farm disconnection and line breakage faults. Based on the line fault probability distribution, known scenarios of line fault states with a probability higher than the predetermined probability are selected. Based on the novel power system recovery optimization model, the optimal recovery decision for each scenario is solved, and a pre-optimized decision set is established. We construct decision indicators with two dimensions: efficiency and risk, for scenarios where the state of line faults is unknown. The efficiency dimension focuses on the efficiency of load recovery, while the risk dimension focuses on the negative impacts of wind farm disconnection from the grid and line outages. A pre-optimized guided Monte Carlo tree search model is constructed. In the selection and expansion phase, operations that match the current state in the pre-optimized decision set are given priority as expansion candidates. In the simulation phase, a directed in-depth simulation strategy that considers the probability of line failure is formulated. In the backtracking phase, the decision index is used as the reward value, and the number of visits and cumulative value of the search tree nodes are updated through backpropagation. The Monte Carlo tree search model is used to search online and output recovery decisions in a rolling manner. The recovery operation corresponding to the child node with the most visits under the root node is selected as the optimal recovery action for the current time step. After executing the optimal recovery action, the state of the power system is updated. The subsequent recovery decision is determined in a rolling manner in the scenario where the line state is unknown.

2. The novel power system restoration decision-making method considering the impact of unknown faults as described in claim 1, characterized in that, The process of constructing a new power system recovery optimization model that considers the impact of wind farm disconnection and line breakage faults, with generator output, load recovery, and load loss as objective functions, includes: optimizing the recovery process from three perspectives: generator output, load recovery, and load loss, with the objective function being the optimal weighted sum of each scenario. It also involves setting connectivity constraints, component recovery sequence constraints, conventional unit output constraints, wind turbine output constraints, and node power balance constraints, and considering the impact of wind farm disconnection and line breakage faults in the process of setting the constraints.

3. A novel power system restoration decision-making method considering the impact of unknown faults as described in claim 1, characterized in that, The process of selecting known scenarios of line fault states with probabilities higher than predetermined probabilities based on the line fault probability distribution includes: calculating the joint probability of each scenario and selecting the scenarios with the highest joint probabilities. K The scenario constitutes a known scenario of a line fault state with a higher probability than expected. ， parameter K The value is determined based on the scale of the power system and computing resources; The joint probability is calculated based on the fault probability and the probability of each line being in a normal state.

4. A novel power system restoration decision-making method considering the impact of unknown faults as described in claim 1, characterized in that, The process of solving the optimal recovery decision for each scenario and establishing a pre-optimized decision set includes: for each scenario in the known scenario of line fault state with a probability higher than the predetermined probability, based on the established new power system recovery optimization model, the optimal recovery decision for the current recovery period is solved by using the mixed integer programming method. The optimal recovery decisions for each scenario in each period are collected to establish a pre-optimized decision set as a guiding prior.

5. A novel power system restoration decision-making method considering the impact of unknown faults as described in claim 1, characterized in that, The process of constructing decision indicators based on both efficiency and risk dimensions in scenarios where the line fault status is unknown includes: The decision indicators are: ； Among them, the risk indicators of load shedding caused by wind farm disconnection are: ； ； Load-benefit metrics for recovery decisions: ； In the formula, α , β These are the weighting coefficients for load revenue and grid disconnection risk, respectively. For the current decision-making step; i index for bus node N For the set of bus nodes; W For the collection of wind farms in the system; t Index for system recovery period; The future time step length in the MCTS simulation; The representative is the first s Under the simulated trajectory t Periodic wind farm w Whether it is offline or not Offline Not offline; indicator variable The value is 1 when at least one wind farm goes offline; For the first s Under the simulated trajectory t Time period nodes The amount of load that has been restored; For the first s The cumulative amount of unloaded load simulated within the future time step under the simulated trajectory; For the first s The simulated load recovery amount within the future time step under the simulated trajectory The summation of variables at each time step effectively introduces a time-weighted effect; loads restored in earlier time periods continue to supply power in subsequent time periods, contributing more to the cumulative index. It is regarded as a comprehensive performance indicator for measuring the power restoration speed and timeliness of the restoration plan.

6. A novel power system restoration decision-making method considering the impact of unknown faults as described in claim 1, characterized in that, The process of constructing a pre-optimized guided Monte Carlo tree search model includes the following steps: The Monte Carlo tree search model gradually constructs a search tree by repeatedly executing four stages: selection, expansion, simulation, and backtracking. Ultimately, the action with the most visits is selected as the optimal decision. In the selection stage, based on the established pre-optimized decision set, with the current recovery state as the root node and candidate recovery paths as branches, when the number of iterations is within a predetermined range, the search process is in the prior guidance stage, prioritizing actions in the optimal recovery sequence for fault scenarios with a higher probability than the predetermined number. When the number of iterations exceeds the predetermined range, the search process enters the global search stage, selecting based on topological connectivity. During the expansion phase, from the lines that are not yet in operation in the current state, lines that meet the connectivity constraints are selected as the candidate action set, set as the child nodes of the current node, and their upper confidence interval index value is set to infinity. The new node inherits the state information of the parent node and updates the set of lines that are already in operation.

7. A novel power system restoration decision-making method considering the impact of unknown faults as described in claim 6, characterized in that, The process of constructing a pre-optimized guided Monte Carlo tree search model includes: during the simulation phase, a directed in-depth simulation strategy that takes into account the probability sampling of line faults is adopted, and the model is preferentially extended along the frontier direction of the currently restored topology to enhance the topological continuity of the simulation path; if there are no more lines to extend in the current direction, the model is then returned to the global set of available actions for selection. During the backtracking phase, the reward values ​​obtained in the simulation phase are propagated backward along the search path to update the statistical information of each node on the path. The simulation is iterated until the accumulated statistical information can reflect the expected benefits and risks of different recovery paths under failure uncertainty.

8. A novel power system restoration decision-making method considering the impact of unknown faults as described in claim 1, characterized in that, The process of using the Monte Carlo tree search model to search online and output recovery decisions includes: embedding a wind farm disconnection fast judgment device into the Monte Carlo tree search model, which is used to judge the wind farm disconnection status under the candidate recovery operation or recovery path by taking the network topology information, fault location and fault duration after the candidate recovery operation or recovery path is executed as input.

9. A novel power system restoration decision-making method considering the impact of unknown faults as described in claim 1, characterized in that, The process of using the Monte Carlo tree search model to search online and output recovery decisions in a rolling manner includes: taking the candidate recovery operations formed in the simulation phase and the off-grid status of each wind farm as input, substituting them into the new power system recovery optimization model, calculating the decision indicators corresponding to the candidate recovery operations or recovery paths, and using them as the reward value of this simulation to propagate back along the search path, updating the access count and cumulative value of each node in the search tree; after reaching the predetermined number of search iterations or the preset search time limit, selecting the recovery operation corresponding to the child node with the most access count under the root node of the search tree as the optimal recovery action for the current time step, executing the optimal recovery action and synchronously updating the power system status; using the updated system status as the initial state of the root node of the new round of search, performing cyclic execution, and outputting subsequent recovery decisions in scenarios where the line status is unknown in a rolling manner; The entire recovery decision-making process after a major power outage is completed until all components to be restored are put into operation or the recovery termination conditions stipulated by the dispatching authority are met.

10. A novel power system recovery decision system that takes into account the impact of unknown faults, characterized in that, include: The system recovery optimization model construction module is configured to use generator output, load recovery amount and load loss amount as objective functions to construct a new power system recovery optimization model that considers the impact of wind farm disconnection and line failure. The pre-optimized decision set solving module is configured to filter known scenarios of line fault states with a probability higher than a predetermined probability based on the line fault probability distribution, and solve the optimal recovery decision for each scenario based on the new power system recovery optimization model to establish a pre-optimized decision set. The decision indicator construction module is configured to construct decision indicators in two dimensions: efficiency and risk, under the scenario where the line fault status is unknown. The efficiency dimension focuses on the efficiency of load recovery, while the risk dimension focuses on the negative impact of wind farm disconnection from the grid and line outage. The pre-optimized guided Monte Carlo tree search building module is configured to construct a pre-optimized guided Monte Carlo tree search model. In the selection and expansion phase, operations that match the current state in the pre-optimized decision set are given priority as expansion candidates. In the simulation phase, a directed in-depth simulation strategy that considers the probability of line failure is formulated. In the backtracking phase, the decision index is used as the reward value, and the number of visits and cumulative value of the search tree nodes are updated through backpropagation. The online search and rolling decision output module is configured to use the Monte Carlo tree search model to search online and output recovery decisions in a rolling manner. It selects the recovery operation corresponding to the child node with the most visits under the root node as the optimal recovery action for the current time step, and updates the state of the power system after executing the optimal recovery action. It determines the subsequent recovery decision in the scenario where the line state is unknown in a rolling manner.

Images