Wind power equipment fault diagnosis optimization algorithm

Abstract

The invention discloses a wind power equipment fault diagnosis optimization algorithm, and belongs to the field of wind power equipment management, and the optimization algorithm comprises the following specific steps: I, collecting and preprocessing various operation data of wind power equipment; iI, constructing an equipment fault map and carrying out feature extraction on the preprocessed data; iII, performing fault diagnosis and classification on the real-time data based on the equipment fault atlas; iV, optimizing a fault processing scheme according to the diagnosed fault and the type thereof; v, evaluating the probability of fault occurrence and formulating a corresponding early warning mechanism; according to the method, local optimum can be avoided, the effectiveness of a fault processing scheme is improved, the flexibility and adaptability of the scheme are enhanced, a feasible solution can be quickly generated, the method is suitable for large-scale data processing, the accuracy of fault probability evaluation is improved, comprehensive fault occurrence probability analysis is provided, targeted maintenance strategies are helped to be formulated, and the method is suitable for large-scale data processing. Resource configuration can be optimized, potential faults can be responded in time, the downtime of equipment is shortened, and the operation cost is reduced.

Description

Technical Field

[0001] The present invention relates to the field of wind power equipment management, and in particular to a wind power equipment fault diagnosis optimization algorithm. Background Art

[0002] Fault diagnosis of wind power equipment is a key link in the operation and management of wind farms. With the continuous growth of global demand for renewable energy, wind power, as an important clean energy, has gradually expanded its application scope. The reliability and stability of wind power equipment directly affect the power generation efficiency and economic benefits of wind farms. However, during the long-term operation of wind power equipment, it is susceptible to environmental factors, material aging and operational errors, leading to various faults. Timely diagnosis and handling of faults can not only reduce equipment downtime, but also reduce maintenance costs and improve the overall operational efficiency of wind farms; therefore, it is particularly important to invent an optimization algorithm for fault diagnosis of wind power equipment.

[0003] After searching, Chinese patent number CN107133632A discloses a method and system for wind power equipment fault diagnosis. Although this invention overcomes the defects and shortcomings of a single method and improves the accuracy of wind turbine equipment fault diagnosis, it is easy to avoid falling into local optimality, reducing the effectiveness of the fault handling solution, and the flexibility and adaptability of the solution are low; in addition, the existing wind power equipment fault diagnosis optimization algorithm has low accuracy in fault probability assessment and cannot provide a comprehensive fault probability analysis, resulting in a long equipment downtime and increased operating costs; for this reason, we propose a wind power equipment fault diagnosis optimization algorithm. Summary of the invention

[0004] The purpose of the present invention is to solve the defects in the prior art and to propose a wind power equipment fault diagnosis optimization algorithm.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] Wind power equipment fault diagnosis optimization algorithm, the specific steps of the optimization algorithm are as follows:

[0007] Ⅰ. Collect and pre-process various operating data of wind power equipment;

[0008] II. Construct equipment fault maps and extract features from preprocessed data;

[0009] III. Fault diagnosis and classification of real-time data based on equipment fault maps;

[0010] IV. Optimize fault handling solutions based on the diagnosed faults and their types;

[0011] V. Assess the probability of failure and develop corresponding early warning mechanisms.

[0012] As a further solution of the present invention, the specific steps of constructing the equipment fault map in step II are as follows:

[0013] S1.1: Collect information on fault types, fault causes and influencing factors from equipment manuals, operation and maintenance records, and expert experience, and organize them into structured and unstructured data. Then, use NLP technology to extract entities and relationships from the collected information.

[0014] S1.2: Based on the extracted entities, establish the relationship between each group of entities, integrate the extracted entities and relationships into a knowledge graph structure in the form of nodes and edges to construct an equipment fault graph, where nodes represent entities and edges represent the relationship between entities;

[0015] S1.3: Use Bayesian networks to model and reason about the information in the equipment failure map to discover potential failure modes or causes, and regularly maintain the knowledge map based on the continuous collection of new data and the updating of knowledge.

[0016] As a further solution of the present invention, the specific steps of feature extraction in step II are as follows:

[0017] S2.1: Arrange the preprocessed running data into a tensor format that meets the CNN input requirements, load a set of pre-trained CNN models, and then input each set of arranged running data into the CNN model;

[0018] S2.2: The CNN model input layer receives each set of running data and transmits it to the model convolution layer. The convolution layer performs convolution operation on each set of running data through the built-in convolution kernel to extract the local features of each set of running data. After the convolution operation is completed, the extracted local features are nonlinearly activated through the ReLU activation function;

[0019] S2.3: After activation, the processed local features are input into the pooling layer to perform pooling on each group of local features extracted by the convolution layer, reduce the size of the feature map, and retain important features. Then, the convolution, activation, and pooling operations are repeated until all convolution layers and pooling layers are processed.

[0020] S2.4: After feature extraction, the features extracted by the convolution and pooling layers are mapped to the output space using a fully connected layer. The output layer then post-processes the data processed by the fully connected layer and outputs the final running feature information.

[0021] As a further solution of the present invention, the specific steps of fault diagnosis and classification in step III are as follows:

[0022] S3.1: Extract the corresponding operation feature information from the real-time collected operation data, and then integrate each group of information in the equipment fault map into the operation feature information. Then, based on the structure of the Bi-GRU model of the input layer, the bidirectional GRU layer and the output layer, build a diagnosis classification model, and divide the integrated operation feature information into a training set and a test set;

[0023] S3.2: The training set data is input into the diagnostic classification model, and then the model performs forward propagation on the training set data. The bidirectional GRU layer performs forward and backward propagation on each group of data at the same time to obtain the hidden state representation of each time step. The output of the bidirectional GRU is then sent to the fully connected layer for feature extraction, and the activation function is applied to obtain the classification result and output;

[0024] S3.3: The loss between the output of the diagnostic classification model and the true label is calculated through the cross entropy loss function, and then the gradient of the loss value for each network layer of the diagnostic classification model is calculated through the back propagation algorithm. Based on the Adam optimizer, the diagnostic classification model parameters are adjusted with the goal of minimizing the cross entropy loss function value;

[0025] S3.4: Evaluate the performance of the diagnostic classification model through the validation set, adjust the model hyperparameters according to the evaluation results, and repeatedly train and validate the diagnostic classification model until the model performance reaches the preset indicators. Then, deploy the trained diagnostic classification model to the wind power equipment management platform;

[0026] S3.5: Collect and pre-process the operating data of wind power equipment in real time, extract features corresponding to the real-time data from the knowledge graph, and merge them with the processed real-time data. Input the integrated real-time data into the trained diagnostic classification model for inference, and output the fault diagnosis results. According to the results output by the model, classify the faults and generate a fault diagnosis report, which includes the fault type, cause and recommended treatment measures.

[0027] As a further solution of the present invention, the specific steps of the optimized fault handling solution in step IV are as follows:

[0028] S4.1: Different processing steps or fault states in the fault handling process are taken as nodes, and the connection relationships and costs between different processing steps or fault states are taken as edges. The corresponding graph model is constructed based on the generated nodes and edges, and then the pheromone values ​​and heuristic information of each group of edges are initialized;

[0029] S4.2: Randomly generate multiple groups of individuals starting from the starting point of the graph model, calculate the selection probability of the next node according to the pheromone value and heuristic information of each group of nodes, and select the action path by random selection according to the calculated selection probability until reaching the target node. After each group of individuals completes a path search, the path and the corresponding cost are recorded. After each group of individuals completes the path, the pheromone of each group of paths is updated;

[0030] S4.3: Repeat the path selection and pheromone update until the pheromone change is less than the preset threshold, and output the path with the highest pheromone as the optimal solution, then use the optimal solution as the initial solution, and randomly perturb the initial solution to generate multiple groups of processing solutions;

[0031] S4.4: Treat each group of treatment solutions as a water body and distribute it to each wind turbine area. Update the water body state according to the flow rule, simulate the flow of water between different treatment solutions, and each water body flows to its neighboring water body, and compare it with the target water body state, adjust the flow, evaluate the state of each water body, calculate its corresponding objective function value, and select the water body with the highest objective function value as the next treatment solution;

[0032] S4.5: According to the water flow and status update, repeat the water flow and update until the maximum number of iterations is reached, compare the advantages and disadvantages of different solutions, determine the final fault handling solution, apply the final fault handling solution to the actual wind power equipment, and evaluate its effect in actual fault handling, and update the solution in real time based on the evaluation results.

[0033] As a further solution of the present invention, the specific steps of evaluating the probability of a fault occurring in step V are as follows:

[0034] S5.1: Collect historical data and real-time data, and construct a state space based on the collected data. Then, take the current state of the wind power equipment as the root node, and generate multiple groups of child nodes according to the subsequent operation state of the wind power equipment, and connect the child nodes to the root node with edges;

[0035] S5.2: Starting from the root node, use the UCT selection strategy to select child nodes until the leaf node is reached. Multiple groups of new child nodes are generated based on the operation status of the current node. Random simulation is started from the newly expanded child nodes. Multiple groups of state sequences are generated and the results are recorded by randomly selecting actions until a fault occurs or the preset end condition is reached, and the simulation is stopped;

[0036] S5.3: Feedback the simulation results to the root node, and update the winning number and visiting number of the nodes of the unified path. Repeat the selection, expansion, simulation and backtracking stages until the set number of iterations is reached. Then, based on the information of each group of nodes, calculate the probability of failure. Based on the calculated failure probability, analyze the risk of the equipment and formulate corresponding maintenance or response measures.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] 1. The wind power equipment fault diagnosis optimization algorithm randomly generates multiple groups of individuals starting from the starting point of the graph model, calculates the selection probability of the next node, and selects the action path by random selection. After each group of individuals completes a path search, each group of individuals updates the pheromone of each group of paths, repeats the path selection and pheromone update until the pheromone change is less than the preset threshold, and outputs the path with the highest pheromone as the optimal solution. After that, the optimal solution is used as the initial solution, and random perturbations are performed to generate multiple groups of treatment solutions. Each group of treatment solutions is used as a water body and distributed in each wind power equipment area. The water body state is updated according to the flow rules to simulate the flow of water between different treatment solutions, and each water body moves to its neighbor The water flow is calculated and compared with the target water body state, the flow is adjusted, the state of each water body is evaluated, the corresponding objective function value is calculated, and the water body with the highest objective function value is selected as the next treatment plan. According to the water body flow and state update, the water body flow and update are repeated until the maximum number of iterations is reached, the advantages and disadvantages of different plans are compared, and the final fault handling plan is determined. The final fault handling plan is applied to actual wind power equipment, and its effect in actual fault handling is evaluated. The plan is updated in real time according to the evaluation effect, which can avoid falling into local optimality, improve the effectiveness of the fault handling plan, enhance the flexibility and adaptability of the plan, and can quickly generate feasible solutions, which is suitable for large-scale data processing.

[0039] 2. The wind power equipment fault diagnosis optimization algorithm collects historical data and real-time data, and constructs a state space based on the collected data. Then, the current wind power equipment state is used as the root node, and multiple groups of child nodes are generated according to the subsequent operation state of the wind power equipment, and the child nodes are connected to the root node with edges. Starting from the root node, the UCT selection strategy is used to select child nodes until the leaf node is reached. Multiple groups of new child nodes are generated based on the operation status of the current node. Random simulation is started from the newly expanded child node. By randomly selecting actions, multiple groups of state sequences are generated and the results are recorded until a fault occurs or the preset end condition is reached. The simulation is stopped, and the simulation results are fed back to the root node. The number of wins and the number of visits of the nodes of the unified path are updated, and the selection, expansion, simulation and backtracking stages are repeated until the set number of iterations is reached. After that, the probability of fault occurrence is calculated according to the information of each group of nodes. According to the calculated fault probability, the risk of the equipment is analyzed, and corresponding maintenance or response measures are formulated to improve the accuracy of fault probability assessment, provide comprehensive fault probability analysis, help formulate targeted maintenance strategies, optimize resource allocation, respond to potential faults in a timely manner, reduce equipment downtime, and reduce operating costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0041] Figure 1 This is a flowchart of the wind power equipment fault diagnosis optimization algorithm proposed in the present invention. DETAILED DESCRIPTION

[0042] Example 1

[0043] Reference Figure 1 , wind power equipment fault diagnosis optimization algorithm, the specific steps of the optimization algorithm are as follows:

[0044] Collect and pre-process various operating data of wind power equipment.

[0045] Construct equipment failure maps and perform feature extraction on preprocessed data.

[0046] Specifically, information on fault types, causes and influencing factors is collected from equipment manuals, operation and maintenance records, and expert experience, and organized into structured and unstructured data. Then, NLP technology is used to extract groups of entities and relationships from the collected information. Based on the extracted entities, relationships between groups of entities are established. The extracted entities and relationships are integrated into a knowledge graph structure in the form of nodes and edges to construct an equipment fault graph, in which nodes represent entities and edges represent relationships between entities. Bayesian networks are used for modeling, and the information in the equipment fault graph is inferred to discover potential failure modes or causes. Based on the continuous collection of new data and the updating of knowledge, the knowledge graph is regularly maintained.

[0047] Specifically, the preprocessed running data is organized into a tensor format that meets the CNN input requirements, and then a set of pre-trained CNN models are loaded, and then each group of organized running data is input into the CNN model. The CNN model input layer receives each group of running data and transmits it to the model convolution layer. The convolution layer performs a convolution operation on each group of running data through a built-in convolution kernel to extract local features of each group of running data. After the convolution operation is completed, the extracted local features are nonlinearly activated through the ReLU activation function. After the activation is completed, the processed local features are input into the pooling layer to perform pooling processing on each group of local features extracted by the convolution layer, reduce the size of the feature map, and retain important features. Then, the convolution, activation and pooling operations are repeated until all convolution layers and pooling layers are processed. After feature extraction, the fully connected layer is used to map the features extracted by the convolution and pooling layers to the output space. Then, the output layer post-processes the data processed by the fully connected layer and outputs the final running feature information.

[0048] Fault diagnosis and classification of real-time data based on equipment fault maps.

[0049] Specifically, the corresponding operation feature information is extracted from the operation data collected in real time, and then each group of information in the equipment fault map is integrated into the operation feature information. Then, based on the structure of the Bi-GRU model of the input layer, the bidirectional GRU layer and the output layer, a diagnosis classification model is constructed, and the integrated operation feature information is divided into a training set and a test set. The training set data is input into the diagnosis classification model, and then the model performs forward propagation on the training set data. The bidirectional GRU layer simultaneously performs forward and backward propagation on each group of data to obtain the hidden state representation of each time step. Then, the output of the bidirectional GRU is sent to the fully connected layer for feature extraction, and the activation function is applied to obtain the classification result and output. The cross entropy loss function is used to calculate the loss between the output of the diagnosis classification model and the true label, and then the back propagation algorithm is used to calculate the loss value for the diagnosis classification model. The gradient of each network layer is calculated, and the parameters of the diagnosis classification model are adjusted based on the Adam optimizer with the goal of minimizing the cross entropy loss function value. The performance of the diagnosis classification model is evaluated through the validation set, and the model hyperparameters are adjusted according to the evaluation results. The diagnosis classification model is repeatedly trained and verified until the model performance reaches the preset indicators. The trained diagnosis classification model is then deployed on the wind power equipment management platform to collect and preprocess the operating data of the wind power equipment in real time, extract features corresponding to the real-time data from the knowledge graph, and merge them with the processed real-time data. The integrated real-time data is input into the trained diagnosis classification model for inference, and the fault diagnosis results are output. According to the results output by the model, fault classification is performed, and a fault diagnosis report is generated, which includes the fault type, cause and recommended treatment measures.

[0050] Example 2

[0051] Reference Figure 1 , wind power equipment fault diagnosis optimization algorithm, the specific steps of the optimization algorithm are as follows:

[0052] Optimize fault handling plans based on the diagnosed faults and their types.

[0053] Specifically, different processing steps or fault states in the fault handling process are taken as nodes, and the connection relationships and costs between different processing steps or fault states are taken as edges. A corresponding graph model is constructed based on the generated nodes and edges. Then, the pheromone values ​​and heuristic information of each group of edges are initialized, and multiple groups of individuals are randomly generated starting from the starting point of the graph model. The selection probability of the next node is calculated according to the pheromone values ​​and heuristic information of each group of nodes, and the action path is selected by random selection according to the calculated selection probability until the target node is reached. After each group of individuals completes a path search, the path and the corresponding cost are recorded. After each group of individuals completes the path, the pheromone of each group of paths is updated, and the path selection and pheromone update are repeated until the pheromone change is less than the preset threshold, and the path with the highest pheromone is output as the optimal solution, and then The optimal solution is used as the initial solution, and the initial solution is randomly perturbed to generate multiple groups of treatment solutions. Each group of treatment solutions is regarded as a water body and allocated to each wind power equipment area. The water body state is updated according to the flow rule, and the flow of water between different treatment solutions is simulated. Each water body flows to its adjacent water body and is compared with the target water body state. The flow is adjusted, the state of each water body is evaluated, and the corresponding objective function value is calculated. The water body with the highest objective function value is selected as the next treatment solution. According to the water body flow and state update, the water body flow and update are repeated until the maximum number of iterations is reached. The advantages and disadvantages of different solutions are compared to determine the final fault handling solution. The final fault handling solution is applied to actual wind power equipment, and its effect in actual fault handling is evaluated. The solution is updated in real time based on the evaluation effect.

[0054] Assess the probability of failure and develop corresponding early warning mechanisms.

[0055] Specifically, historical data and real-time data are collected, and a state space is constructed based on the collected data. The current state of the wind power equipment is used as the root node, and multiple groups of child nodes are generated according to the subsequent operating state of the wind power equipment. The child nodes are connected to the root node with edges. Starting from the root node, the UCT selection strategy is used to select child nodes until the leaf node is reached. Multiple groups of new child nodes are generated based on the operating status of the current node. Random simulation is started from the newly expanded child nodes. By randomly selecting actions, multiple groups of state sequences are generated and the results are recorded until a fault occurs or the preset end condition is reached. The simulation is stopped, and the simulation results are fed back to the root node. The number of wins and the number of visits of the nodes of the unified path are updated. The selection, expansion, simulation and backtracking stages are repeated until the set number of iterations is reached. After that, the probability of a fault is calculated based on the information of each group of nodes. According to the calculated fault probability, the risk of the equipment is analyzed, and corresponding maintenance or response measures are formulated.

Claims

1. Wind power equipment fault diagnosis optimization algorithm, characterized in that: The specific steps of the optimization algorithm are as follows: Ⅰ. Collect and pre-process various operating data of wind power equipment; II. Construct equipment fault maps and extract features from preprocessed data; III. Fault diagnosis and classification of real-time data based on equipment fault maps; IV. Optimize fault handling solutions based on the diagnosed faults and their types; V. Assess the probability of failure and develop corresponding early warning mechanisms.

2. The wind power equipment fault diagnosis optimization algorithm according to claim 1 is characterized in that: The specific steps for constructing the equipment fault map described in step II are as follows: S1.1: Collect information on fault types, fault causes and influencing factors from equipment manuals, operation and maintenance records, and expert experience, and organize them into structured and unstructured data. Then, use NLP technology to extract entities and relationships from the collected information. S1.2: Based on the extracted entities, establish the relationship between each group of entities, integrate the extracted entities and relationships into a knowledge graph structure in the form of nodes and edges to construct an equipment fault graph, where nodes represent entities and edges represent the relationship between entities; S1.3: Use Bayesian networks to model and reason about the information in the equipment failure map to discover potential failure modes or causes, and regularly maintain the knowledge map based on the continuous collection of new data and the updating of knowledge.

3. The wind power equipment fault diagnosis optimization algorithm according to claim 2 is characterized in that: The specific steps of feature extraction in step II are as follows: S2.1: Arrange the preprocessed running data into a tensor format that meets the CNN input requirements, load a set of pre-trained CNN models, and then input each set of arranged running data into the CNN model; S2.2: The CNN model input layer receives each set of running data and transmits it to the model convolution layer. The convolution layer performs convolution operation on each set of running data through the built-in convolution kernel to extract the local features of each set of running data. After the convolution operation is completed, the extracted local features are nonlinearly activated through the ReLU activation function; S2.3: After activation, the processed local features are input into the pooling layer to perform pooling on each group of local features extracted by the convolution layer, reduce the size of the feature map, and retain important features. Then, the convolution, activation, and pooling operations are repeated until all convolution layers and pooling layers are processed. S2.4: After feature extraction, the features extracted by the convolution and pooling layers are mapped to the output space using a fully connected layer. The output layer then post-processes the data processed by the fully connected layer and outputs the final running feature information.

4. The wind power equipment fault diagnosis optimization algorithm according to claim 3 is characterized in that: The specific steps of fault diagnosis and classification described in step III are as follows: S3.1: Extract the corresponding operation feature information from the real-time collected operation data, and then integrate each group of information in the equipment fault map into the operation feature information. Then, based on the structure of the Bi-GRU model of the input layer, the bidirectional GRU layer and the output layer, build a diagnosis classification model, and divide the integrated operation feature information into a training set and a test set; S3.2: The training set data is input into the diagnostic classification model, and then the model performs forward propagation on the training set data. The bidirectional GRU layer performs forward and backward propagation on each group of data at the same time to obtain the hidden state representation of each time step. The output of the bidirectional GRU is then sent to the fully connected layer for feature extraction, and the activation function is applied to obtain the classification result and output; S3.3: The loss between the output of the diagnostic classification model and the true label is calculated through the cross entropy loss function, and then the gradient of the loss value for each network layer of the diagnostic classification model is calculated through the back propagation algorithm. Based on the Adam optimizer, the diagnostic classification model parameters are adjusted with the goal of minimizing the cross entropy loss function value; S3.4: Evaluate the performance of the diagnostic classification model through the validation set, adjust the model hyperparameters according to the evaluation results, and repeatedly train and validate the diagnostic classification model until the model performance reaches the preset indicators. Then, deploy the trained diagnostic classification model to the wind power equipment management platform; S3.5: Collect and pre-process the operating data of wind power equipment in real time, extract features corresponding to the real-time data from the knowledge graph, and merge them with the processed real-time data. Input the integrated real-time data into the trained diagnostic classification model for inference, and output the fault diagnosis results. According to the results output by the model, classify the faults and generate a fault diagnosis report, which includes the fault type, cause and recommended treatment measures.

5. The wind power equipment fault diagnosis optimization algorithm according to claim 4 is characterized in that: The specific steps of optimizing the fault handling solution described in step IV are as follows: S4.1: Different processing steps or fault states in the fault handling process are taken as nodes, and the connection relationships and costs between different processing steps or fault states are taken as edges. The corresponding graph model is constructed based on the generated nodes and edges, and then the pheromone values ​​and heuristic information of each group of edges are initialized; S4.2: Randomly generate multiple groups of individuals starting from the starting point of the graph model, calculate the selection probability of the next node according to the pheromone value and heuristic information of each group of nodes, and select the action path by random selection according to the calculated selection probability until reaching the target node. After each group of individuals completes a path search, the path and the corresponding cost are recorded. After each group of individuals completes the path, the pheromone of each group of paths is updated; S4.3: Repeat the path selection and pheromone update until the pheromone change is less than the preset threshold, and output the path with the highest pheromone as the optimal solution, then use the optimal solution as the initial solution, and randomly perturb the initial solution to generate multiple groups of processing solutions; S4.4: Treat each group of treatment solutions as a water body and distribute it to each wind turbine area. Update the water body state according to the flow rule, simulate the flow of water between different treatment solutions, and each water body flows to its neighboring water body, and compare it with the target water body state, adjust the flow, evaluate the state of each water body, calculate its corresponding objective function value, and select the water body with the highest objective function value as the next treatment solution; S4.5: According to the water flow and status update, repeat the water flow and update until the maximum number of iterations is reached, compare the advantages and disadvantages of different solutions, determine the final fault handling solution, apply the final fault handling solution to the actual wind power equipment, and evaluate its effect in actual fault handling, and update the solution in real time based on the evaluation results.

6. The wind power equipment fault diagnosis optimization algorithm according to claim 1, characterized in that: The specific steps for evaluating the probability of failure occurrence described in step V are as follows: S5.1: Collect historical data and real-time data, and construct a state space based on the collected data. Then, take the current state of the wind power equipment as the root node, and generate multiple groups of child nodes according to the subsequent operation state of the wind power equipment, and connect the child nodes to the root node with edges; S5.2: Starting from the root node, use the UCT selection strategy to select child nodes until the leaf node is reached. Multiple groups of new child nodes are generated based on the operation status of the current node. Random simulation is started from the newly expanded child nodes. Multiple groups of state sequences are generated and the results are recorded by randomly selecting actions until a fault occurs or the preset end condition is reached, and the simulation is stopped; S5.3: Feedback the simulation results to the root node, and update the winning number and visiting number of the nodes of the unified path. Repeat the selection, expansion, simulation and backtracking stages until the set number of iterations is reached. Then, based on the information of each group of nodes, calculate the probability of failure. Based on the calculated failure probability, analyze the risk of the equipment and formulate corresponding maintenance or response measures.

Images