A disaster prevention data synchronization method and terminal based on monte carlo tree search

Abstract

The application discloses a disaster prevention data synchronization method and terminal based on Monte Carlo tree search, receives a data synchronization request; selects one from all nodes to be synchronized as a root node according to the data synchronization request, determines a first evaluation value of the node to be synchronized based on the time complexity of data synchronization, the current period weight, the current area weight, the historical disaster occurrence rate of the current area and the historical daily load data, performs Monte Carlo tree search in all nodes to be synchronized from the root node based on the first evaluation value, determines the child nodes of the root node, and adds the child nodes to the planned path of data synchronization; synchronizes the disaster prevention data of each node in the planned path; and the Monte Carlo tree search is combined with a specific application scenario to quickly and accurately determine the node that needs to be synchronized for disaster prevention data, so that the efficiency and real-time performance of disaster prevention data synchronization are ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of disaster prevention data synchronization, and in particular to a disaster prevention data synchronization method and terminal based on Monte Carlo tree search. BACKGROUND

[0002] With the increasing investment of State Grid Corporation in distribution network construction, the demand for analysis of distribution network planning related data is increasing. Due to the individualization and diversification of analysis scenarios, a complete development cycle from customization development to online deployment is required, and relying on traditional BI (business intelligence) presentation technology or mature market products cannot meet the diversified and effective analysis needs of users.

[0003] A common analysis scenario is a disaster prevention and mitigation application scenario, in which various types of disaster prevention data (such as meteorological data, power grid load data, etc.) need to be synchronized, so that timely analysis and prediction of natural disasters can be made based on the synchronized disaster prevention data, in order to achieve the purpose of disaster prevention, mitigation and rescue. Therefore, the real-time performance of disaster prevention data synchronization is particularly important. SUMMARY

[0004] The technical problem to be solved by the present application is to provide a disaster prevention data synchronization method and terminal based on Monte Carlo tree search, which can improve the real-time performance of disaster prevention data synchronization.

[0005] To solve the above technical problems, one technical solution adopted by the present application is:

[0006] A disaster prevention data synchronization method based on Monte Carlo tree search, comprising the steps of:

[0007] S1, receiving a data synchronization request;

[0008] S2, selecting one of all nodes to be synchronized as a root node according to the data synchronization request, determining a first evaluation value of the node to be synchronized based on the time complexity of data synchronization, the current period weight, the current region weight, the historical disaster occurrence rate of the current region and the historical daily load data, performing Monte Carlo tree search in all nodes to be synchronized from the root node based on the first evaluation value, determining the child nodes of the root node, and adding the child nodes to the planning path of data synchronization;

[0009] S3, synchronizing the disaster prevention data of each node in the planning path.

[0010] To solve the above technical problems, another technical solution adopted by the present application is:

[0011] The application discloses a disaster prevention data synchronization terminal based on Monte Carlo tree search, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein each step in the disaster prevention data synchronization method based on Monte Carlo tree search is realized when the processor executes the computer program.

[0012] The application has the advantages that the synchronization of disaster prevention data is realized based on Monte Carlo tree search, and when the Monte Carlo tree search is performed, the first evaluation of the node to be synchronized is determined based on the time complexity of data synchronization, the current period weight, the current area weight, the historical disaster occurrence rate of the current area and the historical daily load data in combination with the disaster prevention and mitigation application scene, the Monte Carlo tree search is performed from the root node in all the nodes to be synchronized based on the first evaluation, the child node of the root node is determined, and the child node is added to the planned path of data synchronization, so that the node needing the disaster prevention data synchronization is quickly and accurately determined by using the Monte Carlo tree search in combination with the specific application scene, and the efficiency and real-time performance of the disaster prevention data synchronization are ensured. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 A step flow chart of a disaster prevention data synchronization method based on Monte Carlo tree search of the application embodiment;

[0014] Figure 2 A structural schematic diagram of a disaster prevention data synchronization terminal based on Monte Carlo tree search of the application embodiment;

[0015] Figure 3 A schematic diagram of one iteration in the Monte Carlo tree search of the application embodiment. DETAILED DESCRIPTION

[0016] In order to describe the technical content, the achieved purposes and effects of the application in detail, the following will be described in combination with the embodiments and the drawings.

[0017] Please refer to Figure 1 A disaster prevention data synchronization method based on Monte Carlo tree search, comprising the following steps:

[0018] S1, receiving a data synchronization request;

[0019] S2, selecting one from all the nodes to be synchronized as a root node according to the data synchronization request, determining a first evaluation of the node to be synchronized based on the time complexity of data synchronization, the current period weight, the current area weight, the historical disaster occurrence rate of the current area and the historical daily load data, performing the Monte Carlo tree search from the root node in all the nodes to be synchronized based on the first evaluation, determining the child node of the root node, and adding the child node to the planned path of data synchronization.

[0020] S3, synchronizing disaster prevention data of each node in the planned path.

[0021] From the above description, the beneficial effects of the present application are that: based on Monte Carlo tree search to realize the synchronization of disaster prevention data, when performing Monte Carlo tree search, combining the disaster prevention and mitigation application scene, determining the first evaluation of the node to be synchronized based on the time complexity of data synchronization, the current period weight, the current area weight, the historical disaster occurrence rate of the current area and the historical daily load data, and performing Monte Carlo tree search from the root node in all nodes to be synchronized according to the first evaluation, determining the child node of the root node, and adding the child node to the planned path of data synchronization, combining the specific application scene to quickly and accurately determine the node that needs to be synchronized by using Monte Carlo tree search, thereby ensuring the efficiency and real-time performance of disaster prevention data synchronization.

[0022] Further, the step S3 further includes the step of:

[0023] Recording the execution time of each node in the planned path for data synchronization;

[0024] Dynamically adjusting the current period weight corresponding to each node according to the execution time of each node in the planned path.

[0025] From the above description, after the execution of the data synchronization, the execution time of each node in the data synchronization process is recorded as the basis for the next path planning, and the current period weight of each node is dynamically adjusted through the execution time of each node in the data synchronization process, so that the first evaluation of each node can be dynamically adjusted in real time according to the actual synchronization process, and the reliability and accuracy of the selected child node in the Monte Carlo tree search process are further ensured.

[0026] Further, the dynamically adjusting the current period weight corresponding to each node includes:

[0027] The current period weight = the current period weight * (the total data synchronization time - the execution time of the current node / the total data synchronization time).

[0028] From the above description, the current period weight of the node is dynamically adjusted based on the total data synchronization time and the synchronization execution time of the current node, the longer the execution time of the current node, the smaller the current period weight, and the shorter the execution time of the current node, the larger the current period weight, so that the node with faster synchronization speed can get a larger first evaluation in the Monte Carlo tree search process, thereby being more likely to be selected for data synchronization, thereby improving the overall synchronization efficiency and ensuring the real-time requirement of data synchronization.

[0029] Further, the first estimate value is used to determine the UCB value of each child node of the root node, and the child node with the maximum UCB value is determined as the child node of the root node.

[0030] S21, determining the root node as an initial node, and checking nodes downward from the initial node in all the nodes to be synchronized, for a checked node, judging whether the checked node has ended, if yes, determining a second estimate value of the checked node according to a synchronization time of the checked node, and performing step S24, if no, judging whether the checked node is expanded completely, if yes, determining a UCB value of each child node of the checked node according to a first estimate value of each child node, determining a child node with a maximum UCB value as the initial node, and returning to perform the step of checking nodes downward from the initial node in all the nodes to be synchronized, if no, determining an unexpanded action of the checked node, and performing step S22;

[0031] S22, creating a new node in the search tree, the new node being a child node of the checked node, and a state of the child node being a state after the unexpanded action is performed on the checked node;

[0032] S23, performing random simulation on the created child node to determine a second estimate value of the child node;

[0033] S24, updating scores of all nodes on a path from the root node to the checked node according to the second estimate value, and returning to perform step S21 until a preset iteration number or a preset iteration time is reached;

[0034] S25, determining a node with a highest score under the root node as the child node of the root node.

[0035] As can be seen from the above description, the complete Monte Carlo tree search process is realized through selection, expansion, simulation and backtracking, and in the Monte Carlo tree search process, the UCB value is determined adaptively according to the applied scene based on the time complexity of the node to be synchronized through data synchronization, the current period weight, the current area weight, the historical disaster occurrence rate of the current area and the historical daily load data, the node most urgently needing to be expanded is selected according to the maximum UCB value, and finally the node with the highest score is selected as the child node of the root node, thereby ensuring the rationality of the child node obtained through the Monte Carlo tree search.

[0036] Further, the UCB value of the node i is: i

[0037]

[0038] ​In the formula, v i represents the first estimate value of node i, C represents an adjustable bias parameter, n i represents the number of times that node i is accessed, and N represents the number of times that the parent node of node i is accessed.

[0039] As described above, the UCB value considers both the first estimate value of the node and the number of times that the node is accessed and the number of times that the parent node of the node is accessed, and an adjustable bias parameter is added to ensure the rationality and controllability of the calculated UCB value.

[0040] Further, the first estimate value v i of node i corresponds to the following expression:

[0041] v i = time complexity of data synchronization of node i + current period weight of node i + current area weight of node i + historical disaster occurrence rate of the current area of node i + historical daily load data of node i.

[0042] Further, the step S23 comprises:

[0043] If the current node has synchronized data, two historical synchronization times and one random number are randomly selected from the historical data, and three-point estimation is performed according to the two historical synchronization times and the one random number to determine the first estimate value of the child node. If the current node has not synchronized data, three random numbers are selected, and three-point estimation is performed according to the three random numbers to determine the first estimate value of the child node.

[0044] As described above, according to whether the current node has synchronized data, if the current node has synchronized data, two historical synchronization times and one random number are randomly selected from the historical data, and three-point estimation is performed according to the two historical synchronization times and the one random number to determine the first estimate value of the child node. If the current node has not synchronized data, three random numbers are selected, and three-point estimation is performed according to the three random numbers to determine the first estimate value of the child node. This ensures the reliability of the random simulation.

[0045] Further, in the all nodes to be synchronized, starting from the initial node, the nodes are checked downwards. For the checked node, the state of the checked node is obtained. If the state of the checked node is not planned, the step of determining whether the checked node has ended is executed. If the state of the checked node is a skip state or a planned state, the checked node is skipped, the next node is checked, and the step of obtaining the state of the checked node is returned.

[0046] As described above, when the nodes are checked downwards from the initial node, the state of the node to be checked is first obtained. Only when the state of the node to be checked is not planned, the subsequent checking steps are further executed, thereby further improving the efficiency of the Monte Carlo tree search.

[0047] Further, for the inspected node, if the inspected node is currently in the disaster prevention data collection process, the current area weight of the inspected node is reduced, and the current iteration number is determined, if the current iteration number reaches a first threshold value, the state of the inspected node is set to a skip state, the inspected node is placed at the end of the planned path after the iteration is completed, and the disaster prevention data collection is performed on the inspected node again after the disaster prevention data synchronization is completed.

[0048] From the above description, if the inspected node is in the disaster prevention data collection process during the Monte Carlo tree search process, the inspected node is downgraded, the corresponding current area weight is reduced, and after multiple iterations, if the inspected node is still in the disaster prevention data collection process, the state of the inspected node is directly set to a skip state, and the inspected node is placed at the end of the planned path, and the data collection is performed again after the synchronization is completed. According to the real-time situation of each node, the first evaluation value is dynamically adjusted, and the state is changed, so as to further improve the efficiency of the Monte Carlo tree search and ensure the real-time performance of the data synchronization.

[0049] Please refer to Figure 2 A disaster prevention data synchronization terminal based on Monte Carlo tree search, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements each step of the above-mentioned disaster prevention data synchronization method based on Monte Carlo tree search when executing the computer program.

[0050] Embodiment one

[0051] Please refer to Figure 1 A disaster prevention data synchronization method based on Monte Carlo tree search, comprising the steps of:

[0052] S1, receiving a data synchronization request;

[0053] Specifically, the synchronization service master node can receive the data synchronization request from the request end at regular intervals, and the synchronization service master node starts to plan the path of data synchronization after receiving the data synchronization request;

[0054] S2, selecting one from all nodes to be synchronized as a root node according to the data synchronization request, determining the first evaluation value of the node to be synchronized based on the time complexity of data synchronization, the current period weight, the current area weight, the historical disaster occurrence rate of the current area, and the historical daily load data, performing Monte Carlo tree search in all nodes to be synchronized from the root node based on the first evaluation value, determining the child node of the root node, and adding the child node to the planned path of data synchronization;

[0055] The data of each region can be taken as a synchronization node to guarantee the data integrity of each region, and the data to be synchronized includes meteorological data, power grid data, daily load data and the like;

[0056] In the root node selection, in an optional embodiment, one node can be randomly selected from all the nodes to be synchronized, and in another optional embodiment, a key region can be determined in a peak electricity consumption season or in a case of a large meteorological warning, and a root node is selected from the key region, so as to preferentially guarantee the integrity, real-time and availability of the disaster prevention data of the key region in each key period, and the data consistency before the day can be achieved at the end of the day, so as to better achieve the purpose of disaster prevention, disaster reduction and disaster relief;

[0057] S3, synchronizing the disaster prevention data of each node in the planning path, in an optional embodiment, whether to start multi-threaded data synchronization can be decided according to the number of online users and the server load of the system, so as to further improve the real-time of data synchronization;

[0058] In step S2, the Monte Carlo tree search is performed from the root node in all the nodes to be synchronized based on the first estimate value, and the child nodes of the root node include:

[0059] S21, selection stage (Selection):

[0060] The root node is determined as an initial node, and the nodes are checked from the initial node in all the nodes to be synchronized, in the starting stage, the search tree has only one node, that is, the situation to be decided, that is, in the selection stage, the root node is selected, that is, the situation R to be decided is selected to a node N which needs to be expanded most urgently, and the situation R is the first node to be checked in each iteration;

[0061] Each node in the search tree contains three basic information: a situation represented, a number of times visited and a score, and the situation includes the following: to be collected, collected operation completed and to be synchronized, and each node includes the following states: planned, not planned and skipped;

[0062] The to-be-collected includes to-be-collected meteorological data and power grid data, and the collected operation includes daily load data operation of the power grid;

[0063] For the checked node, there are three possibilities:

[0064] The first kind: all the possible actions of the node have been expanded, wherein all the possible actions include the paths from the current node to all the nodes in the un-planned and un-skipped states;

[0065] The second kind: the node has actions that have not been expanded;

[0066] The third kind: the node has been finished, i.e. each node in the path from the root node to the node has reached global data consistency;

[0067] Therefore, for the three possible inspected nodes:

[0068] If the inspected node has been finished, i.e. the inspected node is the third kind, at this time each node in the path from the root node to the node has completed synchronization, then the second estimate value of the inspected node is determined according to the synchronization time of the inspected node, i.e. the inverse number of the execution time of the data synchronization of the inspected node is directly taken as the second estimate value, and the second estimate values of other nodes in the path from the root node to the node are also the inverse numbers of the execution times of the data synchronization of the respective nodes, and step S24 is directly executed from the inspected node; if not, it is determined whether the inspected node is completely expanded, if yes, i.e. the inspected node is the first kind, then the UCB value of each child node of the inspected node is determined according to the first estimate value corresponding to each child node, the child node with the largest UCB value is determined as the initial node, and the step of checking the nodes from the initial node to the bottom in all the nodes to be synchronized is returned to execute, i.e. one child node with the largest UCB value is found to continue to be checked, and iteration is repeated downwards, if not, i.e. the inspected node is the second kind, then it is determined that the inspected node has not been expanded, and step S22 is executed, for example: if the inspected situation still exists and the child node has not been expanded (for example, a node has 20 actions, but only 19 child nodes are created in the search tree), then the node is considered as the target node N of this iteration, and the action A that has not been expanded by N is found, and step S22 is executed;

[0069] Wherein, the UCB value of node i is i The value is:

[0070]

[0071] In the formula, v i represents the first estimate value of node i, C represents a bias parameter that can be adjusted, n i represents the number of times that node i is accessed, and N represents the number of times that the parent node of node i is accessed;

[0072] The first estimate value v i of node i corresponds to:

[0073] v i= time complexity of data synchronization of node i + current period weight of node i + current area weight of node i + historical disaster occurrence rate of current area of node i + historical daily load data of node i;

[0074] The time complexity of data synchronization of node i is the inverse of the time value in milliseconds consumed by node i in performing data synchronization. The current area weight is determined according to the priority of each area. If the priority is higher, the area weight is larger.

[0075] In the first estimation calculation, each parameter is normalized before being added. One normalization method is as follows:

[0076] y = log 10 (x), that is, the logarithmic function conversion with base 10. x and y are the values before and after conversion, respectively.

[0077] In the selection stage, the number of visits of each checked node is increased by one.

[0078] S22, Expansion:

[0079] After the selection stage ends, if the selected node is the second case, the target node N and one unexpanded action A existing in the target node N are found at the end of the selection stage. At this time, a new node is created in the search tree, which is a child node Nn of the checked node. The situation of the child node Nn is the situation after the checked node N performs the unexpanded action A.

[0080] S23, Simulation:

[0081] Random simulation is performed on the created child node to determine the second estimation of the child node.

[0082] Specifically, the step S23 includes:

[0083] It is judged whether the current node Nn has synchronized data. If yes, two historical synchronization times and one random number are randomly selected from the historical data. Three-point estimation is performed according to the two historical synchronization times and the one random number to determine the second estimation of the child node. If no, three random numbers are selected. Three-point estimation is performed according to the three random numbers to determine the second estimation of the child node.

[0084] The second estimation value of the child node is determined by adopting the beta distribution three-point estimation, and the data synchronization process is a cyclic iteration process. After each data synchronization is completed, the synchronization time of the execution synchronization is recorded, and the execution time of each data synchronization is taken as historical data. If the current node has synchronized data, two historical synchronization times are obtained from the historical data, one of which is taken as the best synchronization time, and the other is taken as the most possible synchronization time. Then, a random number of the same order is selected, which is taken as the worst synchronization time. Finally, the second estimation value of the child node is obtained as (the best synchronization time + 4*the most possible synchronization time + the worst synchronization time) / 6. If there is no synchronized data, three random numbers of the same order are selected, the first one is taken as the best synchronization time, the second one is taken as the most possible synchronization time, and the third one is taken as the worst synchronization time. Then, the above-mentioned beta distribution three-point estimation is performed.

[0085] In the embodiment, the time complexity of the node data synchronization can be selected to represent the synchronization time.

[0086] If the checked node has multiple unexpanded actions, the node is expanded and simulated in sequence until the expansion is complete.

[0087] S24, backpropagation:

[0088] If the checked node is the third case, the scores of all nodes on the path from the root node to the checked node are directly updated according to the global data consistent situation. Specifically, the second estimation value of the checked node is taken as its score, and the score of the parent node of the checked node is the average of the second estimation values of all child nodes of the parent node. In this way, the scores of all nodes on the path are updated from bottom to top until the root node. If the checked node is the second case, the scores of all nodes on the path from the root node to the checked node are updated according to the simulation result after the simulation is completed, and the step S21 is returned to be executed until a preset iteration number or a preset iteration time is reached.

[0089] Specifically, the score of the node at the last level is set to the second estimation value obtained by simulation, and the score of the node at the previous level, i.e., the parent node of the node at the last level, is the average of the second estimation values of all corresponding actionable actions under the parent node, i.e., the average of the second estimation values of all child nodes of the parent node. The scores of the nodes are updated from bottom to top until the root node.

[0090] S25, the node with the highest score under the root node is determined as the child node of the root node.

[0091] Each iteration will expand the search tree, and the size of the search tree will increase with the increase of the number of iterations. The search is ended after a certain number of iterations or time, and the child node with the highest score under the root node is selected as the result of this decision, as shown in Figure 3 FIG. 1 is a schematic diagram of one iteration;

[0092] The number of nodes in the planned path can be dynamically adjusted according to actual needs. The iteration is continuously performed according to the determined number of nodes until the planned path meets the determined number of nodes. After determining the result of this decision, the child node with the highest score can be used as the root node, and then step S21 is returned to be executed until the planned path meets the determined number of nodes.

[0093] Embodiment Two

[0094] The embodiment further limits that the step S3 further includes the following steps:

[0095] Recording the execution time of data synchronization of each node in the planned path;

[0096] Dynamically adjusting the current period weight corresponding to each node in the planned path according to the execution time of each node in the planned path;

[0097] Specifically, the dynamically adjusting the current period weight corresponding to each node includes:

[0098] Current period weight = current period weight * (total data synchronization time - execution time of the current node / total data synchronization time).

[0099] Embodiment Three

[0100] The embodiment further limits that:

[0101] In the all nodes to be synchronized, the nodes are checked from the initial node. For the checked node, the state of the checked node is obtained. If the state of the checked node is not planned, it is judged whether the checked node has ended. If the state of the checked node is a skip state or a planned state, the checked node is skipped, the next node is checked, and the step of obtaining the state of the checked node is returned to be executed.

[0102] If the checked node is currently in the disaster prevention data collection process, the current regional weight of the checked node is reduced, and the current iteration number is determined.

[0103] Embodiment four

[0104] Please refer to Figure 2 A disaster prevention data synchronization terminal based on Monte Carlo tree search, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements each step of the disaster prevention data synchronization method based on Monte Carlo tree search in any one of embodiments one to three when executing the computer program.

[0105] In summary, the disaster prevention data synchronization method and terminal based on Monte Carlo tree search provided by the present application realize the synchronization of disaster prevention data based on Monte Carlo tree search. When performing Monte Carlo tree search, the first evaluation of the node to be synchronized is determined based on the time complexity of data synchronization, the current period weight, the current regional weight, the historical disaster occurrence rate of the current region and the historical daily load data in combination with the disaster prevention and mitigation application scenario. According to the first evaluation, Monte Carlo tree search is performed from the root node in all nodes to be synchronized to determine the child node of the root node, and the child node is added to the planning path of data synchronization. In combination with the specific application scenario, Monte Carlo tree search is used to quickly and accurately determine the node that needs to be synchronized for disaster prevention data. During the Monte Carlo tree search process, the parameters involved in the first evaluation are dynamically adjusted, which can realize the relative real-time synchronization of disaster prevention data in each region. Partial complete real-time data is used to generate local disaster prevention warning data, thereby ensuring the efficiency and real-time performance of disaster prevention data synchronization.

[0106] The above description is only an embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent transformation or direct or indirect application in related technical fields based on the content of the specification and drawings is also included in the patent protection scope of the present application.

Claims

1. A disaster prevention data synchronization method based on Monte Carlo tree search, characterized by, The method comprises the steps of: S1, receiving a data synchronization request; S2, selecting one node as a root node from all nodes to be synchronized according to the data synchronization request, determining a first evaluation value of the node to be synchronized based on the time complexity of data synchronization, the current period weight, the current area weight, the historical disaster occurrence rate of the current area and the historical daily load data, performing a Monte Carlo tree search in all nodes to be synchronized starting from the root node based on the first evaluation value, determining a child node of the root node, and adding the child node to a planned path of data synchronization; S3, synchronizing the disaster prevention data of each node in the planned path; The step of performing a Monte Carlo tree search in all nodes to be synchronized starting from the root node based on the first evaluation value and determining a child node of the root node comprises: S21, determining the root node as an initial node, and checking nodes downward from the initial node in all nodes to be synchronized, for a checked node, judging whether the checked node has ended, if yes, determining a second evaluation value corresponding to the checked node according to a synchronization time of the checked node, and performing step S24, if not, judging whether the checked node is fully expanded, if yes, determining a UCB value corresponding to each child node of the checked node according to a first evaluation value corresponding to each child node, determining a child node with a maximum UCB value as an initial node, and returning to perform the step of checking nodes downward from the initial node in all nodes to be synchronized, if not, determining an unexpanded action of the checked node, and performing step S22; S22, creating a new node in a search tree, the new node being a child node of the checked node, and a situation of the child node being a situation after the checked node performs the unexpanded action; S23, performing random simulation on the created child node to determine a second evaluation value of the child node; S24, updating scores of all nodes on a path from the root node to the checked node according to the second evaluation value, and returning to perform step S21 until a preset iteration number or a preset iteration time is reached; S25, determining a node with a highest score under the root node as a child node of the root node.

2. The disaster prevention data synchronization method based on Monte Carlo tree search according to claim 1, wherein, The method further comprises the steps of: Recording execution times of data synchronization of each node in the planned path; Dynamically adjusting a current period weight corresponding to each node in the planned path according to the execution time of the node. 3.The disaster prevention data synchronization method based on Monte Carlo tree search according to claim 2, wherein, The step of dynamically adjusting the current period weight corresponding to each node comprises: Current period weight = current period weight * (total data synchronization time - execution time of the current node / total data synchronization time).

4. The disaster prevention data synchronization method based on Monte Carlo tree search according to claim 1, wherein, Node i's Value is: wherein Ci represents the first estimate of node i, C represents an adjustable bias parameter, Ni represents the number of times node i is visited, N represents the number of times the parent of node i is visited.

5. The disaster prevention data synchronization method based on Monte Carlo tree search according to any one of claims 1 to 4, characterized in that, First estimate corresponding to node i is represented as follows: = time complexity of node i data synchronization + current period weight of node i + current area weight of node i + historical disaster occurrence rate of current area of node i + historical daily load data of node i.

6. The disaster prevention data synchronization method based on Monte Carlo tree search according to claim 1, wherein, The step S23 comprises: If the current node has synchronized data, two historical synchronization times and one random number are randomly selected from the historical data, and a first estimate of the child node is determined based on the two historical synchronization times and the one random number. If the current node has not synchronized data, three random numbers are selected, and a first estimate of the child node is determined based on the three random numbers.

7. The disaster prevention data synchronization method based on Monte Carlo tree search according to claim 1, characterized in that, In the all nodes to be synchronized, starting from the initial node, the state of the checked node is obtained, and if the state of the checked node is not planned, it is determined whether the checked node has ended, if the state of the checked node is a skip state or a planned state, the checked node is skipped, the next node is checked, and the state of the checked node is obtained. 8.The disaster prevention data synchronization method based on Monte Carlo tree search according to claim 7, wherein, If the checked node is currently in the disaster prevention data collection process, the current regional weight of the checked node is reduced, and the current iteration number is determined. If the current iteration number reaches a first threshold, the state of the checked node is set to a skip state, the checked node is placed at the end of the planned path after the iteration ends, and the disaster prevention data collection of the checked node is performed after the disaster prevention data synchronization ends. 9.A disaster prevention data synchronization terminal based on Monte Carlo tree search, comprising a memory, a processor, and a computer program stored on the memory and capable of running on the processor, characterized in that, The processor executes the computer program to implement each step of the disaster prevention data synchronization method based on the Monte Carlo tree search according to any one of claims 1-8.

Images