Power distribution network recovery success rate prediction method and system

Abstract

The invention relates to the technical field of power distribution network recovery, and discloses a power distribution network recovery success rate prediction method and system, and the method comprises the steps: collecting the operation data of a power distribution network; constructing a Bayesian network, defining nodes according to characteristics of the power distribution network, determining a connection relation of edges, and constructing a conditional probability table; the method comprises the following steps: constructing a recovery strategy through a Markov decision process for dynamic optimization, training the recovery strategy through a reinforcement learning algorithm, introducing a Monte Carlo tree search algorithm to search an optimal solution for the recovery strategy, calculating the probability that the power distribution network is recovered to a target state, and generating success rate prediction results of different recovery strategies. According to the invention, the interpretability and adaptability of the model are improved; dynamic optimization of a recovery strategy is realized, and different fault scenes and system states can be adapted; the calculation efficiency is improved, factors such as system topology, load characteristics and new energy output are comprehensively considered, the prediction accuracy and practicability are improved, the system recovery efficiency is improved, and the power failure loss is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of power distribution network restoration, and in particular to a method and system for predicting a success rate of power distribution network restoration. Background Art

[0002] With the high proportion of renewable energy connected to the distribution network, the system recovery process has become more complex and uncertain. Traditional system recovery methods are mainly based on deterministic models and expert experience, which makes it difficult to accurately predict the success rate of recovery strategies, resulting in low recovery efficiency. At the same time, they lack sufficient consideration of the randomness of renewable energy output. Existing models mostly use deterministic parameters and cannot accurately reflect the fluctuation characteristics of renewable energy generation. The impact of system topology and load characteristics on the recovery process is ignored. Most methods only focus on the power supply side and do not fully consider the network structure and load dynamic characteristics. Summary of the invention

[0003] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0004] In view of the above existing problems, the present invention is proposed. Therefore, the present invention provides a method for predicting the success rate of power distribution network restoration to solve the above problems.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] In a first aspect, the present invention provides a method for predicting a success rate of power distribution network restoration, comprising:

[0007] Collect distribution network operation data;

[0008] Construct a Bayesian network, define nodes according to the characteristics of the distribution network, determine the connection relationship of the edges, and construct a conditional probability table;

[0009] Construct a recovery strategy through Markov decision process for dynamic optimization, defining the state space, action space, and reward function;

[0010] The recovery strategy is trained by a reinforcement learning algorithm, and a Monte Carlo tree search algorithm is introduced to find the optimal solution for the recovery strategy;

[0011] Based on the distribution network operation data and the recovery strategy, the probability of the distribution network recovering to the target state is calculated through the Bayesian network reasoning and Monte Carlo tree search algorithm, and the success rate prediction results of different recovery strategies are generated.

[0012] As a preferred solution of the method for predicting the success rate of distribution network restoration described in the present invention, the construction of the Bayesian network includes:

[0013] Define nodes according to the characteristics of the distribution network and define the node set V of the Bayesian network = {V1, V2, ..., V n}, the node set includes a system topology node, a power node, a load node and a system status node;

[0014] Determine edges based on the nodes, that is, determine dependencies between nodes, and construct a directed acyclic graph G=(V,E), where E is an edge set;

[0015] Construct a conditional probability table for each node V i , define the conditional probability distribution P(V i |Pa(V i )), where Pa(V i ) indicates V i The parent node set of

[0016] Based on the conditional independence assumption, the joint probability distribution is expressed as:

[0017] P(V1,V2,...,V n )=∏ i P(V i |Pa(V i )).

[0018] As a preferred solution of the method for predicting the success rate of distribution network restoration described in the present invention, the method of constructing a restoration strategy through a Markov decision process for dynamic optimization includes:

[0019] Define the state space S, which defines the configuration in the distribution network;

[0020] Define the action space A, including the control operations of the distribution network;

[0021] The state transition probability P(s'|s,a) is calculated through the Bayesian network model, which represents the probability of transitioning to state s' after taking action a in state s;

[0022] Define the reward function R(s,a), expressed as:

[0023] R(s,a)=ω1ΔL+ω2ΔE+ω3C

[0024] Among them, ΔL represents the restored load, ΔE represents the incremental utilization of new energy, C represents the operating cost, and ω1, ω2, and ω3 are weight coefficients;

[0025] Define the state value function V(s) and action value function Q(s,a), expressed as:

[0026] V(s)=max a [R*(s,a)+γ∑ s′ P(s′|s,a)V(s′)]

[0027] Q(s,a)=R(s,a)+γ∑ s′ P(s′|s,a)V(s′)

[0028] Here, γ represents the discount factor.

[0029] As a preferred solution of the method for predicting the success rate of power distribution network restoration of the present invention, the training of the restoration strategy includes:

[0030] Initialize the Q table and initialize the Q values ​​of all state-action pairs to random numbers;

[0031] At each time step t, based on the current state s t Select action a t , observe the reward r t and the next state s t+1 , update the Q value;

[0032] A greedy strategy is used to balance exploration and exploitation, randomly selecting actions with probability ε, and selecting the action with the largest current Q value with probability 1-ε;

[0033] Repeat the above steps and stop learning when the Q value converges or the maximum number of iterations is reached.

[0034] As a preferred solution of the method for predicting the success rate of distribution network restoration of the present invention, the method of introducing the Monte Carlo tree search algorithm to find the optimal solution for the restoration strategy includes:

[0035] Starting from the root node, the most promising child nodes are selected according to the tree-tree strategy until a leaf node is reached;

[0036] If the leaf node is not in a terminal state, randomly select an unexpanded child node and add it to the tree;

[0037] Starting from the newly added node, a fast simulation is performed using the default strategy until the termination state is reached;

[0038] Back propagate the simulation results along the search path and update the statistical information of the nodes.

[0039] As a preferred solution of the method for predicting the success rate of distribution network restoration described in the present invention, the tree strategy algorithm is expressed as:

[0040]

[0041] Among them, Xj represents the average reward of node j, n represents the number of visits to the parent node, n j represents the number of visits to node j, and c is the exploration parameter.

[0042] As a preferred solution of the method for predicting the success rate of distribution network restoration of the present invention, the generation of success rate prediction results of different restoration strategies includes:

[0043] Inputting the distribution network operation data and the restoration strategy to be evaluated;

[0044] Distribution network state transfer is performed through trained Q-table or Monte Carlo tree search tree;

[0045] At each time step, the state transition probability is calculated through the Bayesian network, and the probability of recovering to the target state is accumulated;

[0046] Generate success rate prediction results of different restoration strategies, wherein the success rate prediction results include overall restoration success rate, key load restoration probability, new energy utilization rate, and expected restoration time.

[0047] In a second aspect, the present invention provides a distribution network restoration success rate prediction system, comprising:

[0048] A collection module, used to collect distribution network operation data;

[0049] The first construction module is used to construct a Bayesian network, define nodes according to the characteristics of the distribution network, determine the connection relationship of the edges, and construct a conditional probability table;

[0050] The second building block is used to construct a recovery strategy for dynamic optimization through a Markov decision process, defining the state space, action space, and reward function;

[0051] An optimization module, used to train the recovery strategy through a reinforcement learning algorithm and introduce a Monte Carlo tree search algorithm to find an optimal solution for the recovery strategy;

[0052] The output module is used to calculate the probability of the distribution network recovering to the target state based on the distribution network operation data and the recovery strategy, and generate success rate prediction results of different recovery strategies through the Bayesian network reasoning and Monte Carlo tree search algorithm.

[0053] In a third aspect, the present invention provides an electronic device, comprising:

[0054] Memory and processor;

[0055] The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the distribution network restoration success rate prediction method are implemented.

[0056] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the method for predicting the success rate of distribution network restoration.

[0057] Compared with the prior art, the present invention has the following beneficial effects: the present invention uses Bayesian networks to describe the uncertainty in the system recovery process, thereby improving the interpretability and adaptability of the model; dynamically optimizing the recovery strategy through MDP and reinforcement learning, which can adapt to different fault scenarios and system states; improving computing efficiency, introducing the MCTS method, and realizing efficient search in large-scale state space, overcoming the "curse of dimensionality" problem; comprehensively considering multiple factors such as system topology, load characteristics, and new energy output, thereby improving the accuracy and practicality of prediction. The implementation of the present invention can provide reliable decision-making support for distribution network operators, improve system recovery efficiency, and reduce power outage losses, which is of great significance to improving the reliability and resilience of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0059] Figure 1 The figure is a schematic diagram of the overall process of a method for predicting the success rate of distribution network restoration according to an embodiment of the present invention. DETAILED DESCRIPTION

[0060] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0061] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0062] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0063] The present invention is described in detail with reference to the schematic diagram. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0064] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0065] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0066] Reference Figure 1 , as an embodiment of the present invention, provides a method for predicting the success rate of power distribution network restoration, comprising:

[0067] S101, collecting distribution network operation data;

[0068] S102, constructing a Bayesian network, defining nodes according to the characteristics of the distribution network, determining the connection relationship of the edges, and constructing a conditional probability table;

[0069] S103, constructing a recovery strategy through a Markov decision process for dynamic optimization, defining the state space, action space, and reward function;

[0070] S104, training the recovery strategy through a reinforcement learning algorithm, and introducing a Monte Carlo tree search algorithm to find the optimal solution for the recovery strategy;

[0071] S105, based on the distribution network operation data and recovery strategy, the probability of the distribution network recovering to the target state is calculated through Bayesian network reasoning and Monte Carlo tree search algorithm, and the success rate prediction results of different recovery strategies are generated.

[0072] It should be noted that this embodiment aims to establish a probabilistic model that can accurately describe the uncertainty of the recovery process of a distribution network system with a high proportion of renewable energy access, and to design a recovery strategy evaluation framework that can comprehensively consider multiple factors such as system topology, load characteristics, and renewable energy output, so as to achieve dynamic optimization of the recovery strategy to adapt to different fault scenarios and improve recovery efficiency.

[0073] The present invention proposes a distribution network system restoration success rate prediction model based on Bayesian network and Markov decision process (MDP). The model adopts Bayesian network to describe the probabilistic dependency relationship between various factors in the system restoration process, uses MDP to construct a restoration strategy evaluation framework, optimizes the restoration strategy through reinforcement learning algorithm, and introduces Monte Carlo tree search (MCTS) method to improve prediction efficiency and accuracy.

[0074] Specifically, in step S101, data preparation is first performed to collect distribution network operation data, mainly including system topology, equipment parameters, historical fault and recovery data and other information.

[0075] In a preferred embodiment, step S102, constructing a Bayesian network includes:

[0076] Define nodes according to the characteristics of the distribution network and define the node set V of the Bayesian network = {V1, V2, ..., V n}, the node set includes system topology nodes, power supply nodes, load nodes and system status nodes;

[0077] Among them, the system topology node represents the status of key switches and lines, the power supply node includes conventional power supply and new energy output, the load node represents the power supply status of various loads, and the system status node describes the overall recovery degree.

[0078] Through expert knowledge and historical data analysis, edges are determined based on nodes, that is, the dependencies between nodes are determined, and a directed acyclic graph G = (V, E) is constructed, where E is the edge set;

[0079] Construct a conditional probability table for each node V i , define the conditional probability distribution P(V i |Pa(V i )), where Pa(V i ) indicates V i The parent node set of

[0080] Among them, the CPT of new energy nodes can be obtained by fitting historical data, and the CPT of other nodes can be determined by combining expert experience and historical statistical data;

[0081] Based on the conditional independence assumption, the joint probability distribution is expressed as:

[0082] P(V1,V2,...,V n )=∏ i P(V i |Pa(V i )).

[0083] It should be noted that the Bayesian network is a probabilistic graphical model suitable for describing the conditional independence between variables in a complex system. This embodiment uses the Bayesian network to model the distribution network system recovery process to capture the mutual influence between various factors.

[0084] In a preferred embodiment, step S103, constructing a recovery strategy through a Markov decision process for dynamic optimization includes:

[0085] Define the state space S, which defines the configurations in the distribution network, specifically all possible configurations in the system, including switch states, power output, and load power supply conditions;

[0086] Define the action space A, including the control operations of the distribution network, all possible control operations, such as switch operation, power dispatch, etc.;

[0087] The state transition probability P(s'|s,a) is calculated through the Bayesian network model, which represents the probability of transitioning to state s' after taking action a in state s;

[0088] Define the reward function R(s,a), which is the effect of the recovery operation, considering factors such as recovery load and operation cost, and is expressed as:

[0089] R(s,a)=ω1ΔL+ω2ΔE+ω3C

[0090] Among them, ΔL represents the restored load, ΔE represents the incremental utilization of new energy, C represents the operating cost, and ω1, ω2, and ω3 are weight coefficients;

[0091] Define the state value function V(s) and action value function Q(s,a), expressed as:

[0092] V(s)=max a [R*(s,a)+γ∑ s′ P(s′|s,a)V(s′)]

[0093] Q(s,a)=R(s,a)+γ∑ s′ P(s′|s,a)V(s′)

[0094] Here, γ represents the discount factor.

[0095] It should be noted that the Markov decision process (MDP) is a mathematical model that describes sequential decision problems. This embodiment uses MDP to construct a recovery strategy evaluation framework to achieve dynamic optimization.

[0096] In a preferred embodiment, step S103, training the recovery strategy by using a reinforcement learning algorithm includes:

[0097] Initialize the Q table and initialize the Q values ​​of all state-action pairs to small random numbers;

[0098] Perform iterative updates at each time step t based on the current state s t Select action a t , observe the reward r t and the next state s t+1 , update the Q value, expressed as:

[0099] Q(s t ,a t )←Q(s t ,a t )+α[r t +γmax a Q(s t+1 ,a t )-Q(s t ,a t )]

[0100] Among them, α is the learning rate;

[0101] A greedy strategy is used to balance exploration and exploitation, randomly selecting actions with probability ε, and selecting the action with the largest current Q value with probability 1-ε;

[0102] Repeat the above steps and stop learning when the Q value converges or the maximum number of iterations is reached.

[0103] It should be noted that this embodiment uses the Q-learning algorithm to optimize the recovery strategy, the core of which is to learn the optimal action value function Q(s, a) through repeated experiments.

[0104] In a preferred embodiment, step S104, introducing a Monte Carlo tree search algorithm to find an optimal solution for the recovery strategy includes:

[0105] Starting from the root node, the most promising child nodes are selected according to the tree strategy until a leaf node is reached, where the tree strategy can be the UCB1 strategy;

[0106] If the leaf node is not a terminal state, a random unexpanded child node is selected and added to the tree;

[0107] Starting from the newly added node, use the default strategy to perform a quick simulation until the termination state is reached, where the default strategy can be a random strategy, etc.;

[0108] Back propagate the simulation results along the search path and update the statistical information of the nodes.

[0109] The tree strategy algorithm is expressed as:

[0110]

[0111] Among them, X j represents the average reward of node j, n represents the number of visits to the parent node, n j represents the number of visits to node j, and c is the exploration parameter.

[0112] It should be noted that in order to improve the prediction efficiency and accuracy in large-scale systems, this embodiment introduces the MCTS method. MCTS finds a nearly optimal recovery strategy under limited computing resources by selectively expanding the decision tree.

[0113] In a preferred embodiment, step S105, generating success rate prediction results of different recovery strategies includes:

[0114] Input distribution network operation data and restoration strategies to be evaluated;

[0115] Distribution network state transfer is performed through trained Q-table or Monte Carlo tree search tree;

[0116] At each time step, the state transition probability is calculated through the Bayesian network, and the probability of recovering to the target state is accumulated;

[0117] Generate success rate prediction results for different restoration strategies, including overall restoration success rate, key load restoration probability, new energy utilization rate, and expected restoration time.

[0118] It should be noted that the distribution network system recovery success rate prediction model proposed in this embodiment takes maximizing the system recovery success rate as the optimization goal. The model comprehensively considers multiple factors such as system topology, load characteristics, and new energy output, and constitutes a typical sequential decision-making problem. In order to solve this model, this embodiment adopts a method based on reinforcement learning and MCTS.

[0119] Exemplarily, the solution process of this embodiment is as follows:

[0120] System initialization: Collect and preprocess system data, including topology, device parameters, historical fault and recovery data, etc.

[0121] Bayesian network construction: Define the node set V according to the system characteristics, determine the edge connection relationship E, construct a directed acyclic graph G = (V, E), and use historical data and expert knowledge to construct a conditional probability table (CPT).

[0122] MDP framework design: define the state space S, action space A, state transition probability P(s'|s,a) and reward function R(s,a), and initialize the value function V(s) and Q(s,a).

[0123] Reinforcement learning training: Use the Q-learning algorithm to train the recovery strategy, including:

[0124] Step 1: Initialize the Q table

[0125] Step 2. For each training round:

[0126] Initialize system state s;

[0127] For each time step t:

[0128] Use a greedy strategy to select action a;

[0129] Execute action a, observe reward r and new state s';

[0130] Update Q value;

[0131] Until the terminal state or the maximum number of steps is reached;

[0132] Step 3: Repeat step 2 until the Q value converges or the maximum number of rounds is reached;

[0133] MCTS Optimization: Applying MCTS methods to improve search efficiency in large-scale systems, including:

[0134] Step 1: Starting from the root node, repeat the following steps until the computational budget is reached:

[0135] Selection,Use the UCB1 algorithm to select the most potential child node;

[0136] Expand, if the selected node is not a leaf node, add a new child node;

[0137] Simulate, start a random simulation from a new node;

[0138] Backward propagation, updating the statistics of nodes on the search path.

[0139] Success rate prediction, including:

[0140] Enter the initial failure state and the recovery strategy to be evaluated;

[0141] Use the trained Q table or MCTS tree to guide the system state transition;

[0142] At each time step, the state transition probability is calculated using a Bayesian network;

[0143] Cumulatively calculate the probability that the system will return to the target state;

[0144] The result output generates success rate prediction results of different recovery strategies, including: overall recovery success rate, key load recovery probability, new energy utilization rate, and expected recovery time.

[0145] It should be noted that the present invention uses Bayesian networks to describe the uncertainty in the system recovery process, thereby improving the interpretability and adaptability of the model; dynamically optimizing the recovery strategy through MDP and reinforcement learning can adapt to different fault scenarios and system states; improving computing efficiency, introducing the MCTS method, and realizing efficient search in large-scale state space, overcoming the "curse of dimensionality" problem; comprehensively considering multiple factors such as system topology, load characteristics, and new energy output, thereby improving the accuracy and practicality of the prediction. The implementation of the present invention can provide reliable decision-making support for distribution network operators, improve system recovery efficiency, and reduce power outage losses, which is of great significance to improving the reliability and resilience of the distribution network.

[0146] The above is a schematic scheme of a distribution network recovery success rate prediction method of this embodiment. It should be noted that the technical scheme of the distribution network recovery success rate prediction system and the technical scheme of the distribution network recovery success rate prediction method mentioned above belong to the same concept. For details not described in detail in the technical scheme of the distribution network recovery success rate prediction system in this embodiment, please refer to the description of the technical scheme of the distribution network recovery success rate prediction method mentioned above.

[0147] The distribution network restoration success rate prediction system in this embodiment includes:

[0148] A collection module, used to collect distribution network operation data;

[0149] The first construction module is used to construct a Bayesian network, define nodes according to the characteristics of the distribution network, determine the connection relationship of the edges, and construct a conditional probability table;

[0150] The second building block is used to construct a recovery strategy for dynamic optimization through a Markov decision process, defining the state space, action space, and reward function;

[0151] The optimization module is used to train the recovery strategy through the reinforcement learning algorithm and introduce the Monte Carlo tree search algorithm to find the optimal solution for the recovery strategy;

[0152] The output module is used to calculate the probability of the distribution network recovering to the target state based on the distribution network operation data and recovery strategy, through Bayesian network reasoning and Monte Carlo tree search algorithm, and generate the success rate prediction results of different recovery strategies.

[0153] This embodiment further provides an electronic device, which is applicable to the case of predicting the success rate of power distribution network restoration, and includes:

[0154] Memory and processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the method for predicting the success rate of distribution network restoration as proposed in the above embodiment.

[0155] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for predicting the success rate of distribution network restoration proposed in the above embodiment is implemented.

[0156] The storage medium proposed in this embodiment and the method for predicting the success rate of distribution network recovery proposed in the above embodiment belong to the same inventive concept. The technical details not described in detail in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0157] Through the above description of the implementation methods, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ReadOnly, Memory, ROM), random access memory (RandomAccess Memory, RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform the methods of various embodiments of the present invention.

[0158] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for predicting the success rate of distribution network restoration, characterized in that: include: Collect distribution network operation data; Construct a Bayesian network, define nodes according to the characteristics of the distribution network, determine the connection relationship of the edges, and construct a conditional probability table; Construct a recovery strategy through Markov decision process for dynamic optimization, defining the state space, action space, and reward function; The recovery strategy is trained by a reinforcement learning algorithm, and a Monte Carlo tree search algorithm is introduced to find the optimal solution for the recovery strategy; Based on the distribution network operation data and the recovery strategy, the probability of the distribution network recovering to the target state is calculated through the Bayesian network reasoning and Monte Carlo tree search algorithm, and the success rate prediction results of different recovery strategies are generated.

2. The method for predicting the success rate of power distribution network restoration according to claim 1, characterized in that: The construction of the Bayesian network includes: Define nodes according to the characteristics of the distribution network and define the node set V of the Bayesian network = {V1, V2, ..., V n }, the node set includes a system topology node, a power node, a load node and a system status node; Determine edges based on the nodes, that is, determine dependencies between nodes, and construct a directed acyclic graph G=(V,E), where E is an edge set; Construct a conditional probability table for each node V i , define the conditional probability distribution P(V i |Pa(V i )), where Pa(V i ) indicates V i The parent node set of Based on the conditional independence assumption, the joint probability distribution is expressed as: P(V1,V2,...,V n )=∏ i P(V i |Pa(V i ))。 3. The method for predicting the success rate of distribution network restoration according to claim 1 or 2, characterized in that: The dynamic optimization of the recovery strategy constructed by Markov decision process includes: Define the state space S, which defines the configuration in the distribution network; Define the action space A, including the control operations of the distribution network; The state transition probability P(s'|s,a) is calculated through the Bayesian network model, which represents the probability of transitioning to state s' after taking action a in state s; Define the reward function R(s,a), expressed as: R(s,a)=ω1ΔL+ω2ΔE+ω3C Among them, ΔL represents the restored load, ΔE represents the incremental utilization of new energy, C represents the operating cost, and ω1, ω2, and ω3 are weight coefficients; Define the state value function V(s) and action value function Q(s,a), expressed as: V(s)=max a [R*(s,a)+γ∑ s′ P(s′|s,a)V(s′)] Q(s,a)=R(s,a)+γ∑ s′ P(s′|s,a)V(s′) Here, γ represents the discount factor.

4. The method for predicting the success rate of power distribution network restoration according to claim 1, characterized in that: Training the recovery strategy through a reinforcement learning algorithm includes: Initialize the Q table and initialize the Q values ​​of all state-action pairs to random numbers; At each time step t, based on the current state s t Select action a t , observe the reward r t and the next state s t+1 , update the Q value; A greedy strategy is used to balance exploration and exploitation, randomly selecting actions with probability ε, and selecting the action with the largest current Q value with probability 1-ε; Repeat the above steps and stop learning when the Q value converges or the maximum number of iterations is reached.

5. The method for predicting the success rate of distribution network restoration according to claim 4, characterized in that: The introduction of the Monte Carlo tree search algorithm to find the optimal solution for the recovery strategy includes: Starting from the root node, the most promising child nodes are selected according to the tree-tree strategy until a leaf node is reached; If the leaf node is not in a terminal state, randomly select an unexpanded child node and add it to the tree; Starting from the newly added node, a fast simulation is performed using the default strategy until the termination state is reached; Back propagate the simulation results along the search path and update the statistical information of the nodes.

6. The method for predicting the success rate of power distribution network restoration according to claim 5, characterized in that: The tree strategy algorithm is expressed as: Among them, X j represents the average reward of node j, n represents the number of visits to the parent node, n j represents the number of visits to node j, and c is the exploration parameter.

7. The method for predicting the success rate of distribution network restoration according to claim 6, characterized in that: The generation of success rate prediction results of different recovery strategies includes: Inputting the distribution network operation data and the restoration strategy to be evaluated; Distribution network state transfer is performed through trained Q-table or Monte Carlo tree search tree; At each time step, the state transition probability is calculated through the Bayesian network, and the probability of recovering to the target state is accumulated; Generate success rate prediction results of different restoration strategies, wherein the success rate prediction results include overall restoration success rate, key load restoration probability, new energy utilization rate, and expected restoration time.

8. A distribution network restoration success rate prediction system, characterized in that: include, A collection module, used to collect distribution network operation data; The first construction module is used to construct a Bayesian network, define nodes according to the characteristics of the distribution network, determine the connection relationship of the edges, and construct a conditional probability table; The second building block is used to construct a recovery strategy for dynamic optimization through a Markov decision process, defining the state space, action space, and reward function; An optimization module, used to train the recovery strategy through a reinforcement learning algorithm and introduce a Monte Carlo tree search algorithm to find an optimal solution for the recovery strategy; The output module is used to calculate the probability of the distribution network recovering to the target state based on the distribution network operation data and the recovery strategy, and generate success rate prediction results of different recovery strategies through the Bayesian network reasoning and Monte Carlo tree search algorithm.

9. An electronic device, characterized in that: include: Memory and processor; The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the distribution network restoration success rate prediction method described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: It stores computer executable instructions, which, when executed by a processor, can implement the steps of the method for predicting the success rate of distribution network restoration as described in any one of claims 1 to 7.

Images