Intelligent inspection methods, devices, computer equipment and storage media for nuclear power plants

Abstract

This invention relates to the field of nuclear power plant safety operation, and discloses a method, device, computer equipment, and storage medium for intelligent inspection of nuclear power plants. The method includes: acquiring real-time monitoring data within an inspection area through a first interface; acquiring real-time inspection records from an inspection robot through a second interface; processing the real-time monitoring data using a first nuclear power plant inspection strategy evaluation model to obtain a first inspection strategy; processing the real-time inspection records using a second nuclear power plant inspection strategy evaluation model to obtain a second inspection strategy; and determining a target inspection strategy based on the first and second inspection strategies. This invention can improve inspection efficiency, improve inspection quality, and reduce inspection costs.

Description

Technical Field

[0001] This invention relates to the field of safe operation of nuclear power plants, and in particular to a method, device, computer equipment and storage medium for intelligent inspection of nuclear power plants. Background Technology

[0002] Nuclear power plant equipment is diverse, and its operating status is affected by various factors. For example, equipment configuration, usage frequency, ambient temperature, and humidity all influence its operational status. Therefore, the inspection plan for the equipment in the power plant room needs to comprehensively consider various factors and formulate a scientific and reasonable maintenance plan and management scheme. Currently, nuclear power plant power plant room inspection robots can only execute pre-set inspection tasks and cannot dynamically adjust the inspection tasks. This inspection method lacks flexibility and adaptability, and cannot be dynamically adjusted according to the actual situation of the equipment. It requires personnel to adjust the inspection tasks, resulting in low inspection efficiency and a high risk of missed or false inspections. Summary of the Invention

[0003] Therefore, it is necessary to provide a method, device, computer equipment, and storage medium for intelligent inspection of nuclear power plants to address the aforementioned technical problems, so as to improve the efficiency and quality of nuclear power plant inspections.

[0004] A method for intelligent inspection of nuclear power plants includes:

[0005] Real-time monitoring data within the patrol area is obtained through the first interface;

[0006] The real-time inspection records of the inspection robot are obtained through the second interface;

[0007] The real-time monitoring data is processed by the first nuclear power plant inspection strategy evaluation model to obtain the first inspection strategy;

[0008] The real-time inspection records are processed by the second nuclear power plant inspection strategy evaluation model to obtain the second inspection strategy;

[0009] The target inspection strategy is determined based on the first inspection strategy and the second inspection strategy.

[0010] Optionally, acquiring real-time monitoring data within the patrol area through the first interface includes:

[0011] Communication is established with multiple nuclear power plant systems within the inspection area through the first interface;

[0012] Raw monitoring data were obtained from the multiple nuclear power plant systems;

[0013] The original monitoring data is filtered according to preset filtering rules to obtain the real-time monitoring data.

[0014] Optionally, obtaining the real-time inspection records of the inspection robot through the second interface includes:

[0015] The inspection robot collects raw inspection data.

[0016] The robot preprocesses the raw inspection data to obtain the real-time inspection record; the real-time inspection record includes path trajectory, image recognition results, and abnormal event records.

[0017] The real-time inspection records are obtained from the robot terminal using a timed method and / or an event-triggered method.

[0018] Optionally, the first nuclear power plant inspection strategy evaluation model includes a nuclear power plant equipment risk prediction model and a nuclear power plant inspection planning model; the nuclear power plant equipment risk prediction model is a hybrid model based on Bayesian networks and long short-term memory networks;

[0019] The process of processing the real-time monitoring data through the first nuclear power plant inspection strategy evaluation model to obtain the first inspection strategy includes:

[0020] The real-time monitoring data is processed by the nuclear power plant equipment risk prediction model to obtain the predicted status of the nuclear power plant equipment.

[0021] The nuclear power plant inspection planning model is used to process the predicted status of the nuclear power plant equipment to obtain the first inspection strategy.

[0022] Optionally, the step of processing the real-time inspection records through the second nuclear power plant inspection strategy evaluation model to obtain the second inspection strategy includes:

[0023] The action space and state space of the second nuclear power plant inspection strategy evaluation model are set according to the real-time inspection records.

[0024] Multiple candidate inspection paths are planned within the aforementioned action space:

[0025] The reward value of the candidate inspection path in the state space is calculated according to the preset reward function;

[0026] The inspection path with the highest reward value is selected as the second inspection path.

[0027] The second inspection strategy is generated based on the second inspection path.

[0028] Optionally, determining the target inspection strategy based on the first inspection strategy and the second inspection strategy includes:

[0029] Obtain the first multi-dimensional weight data of the first inspection strategy; obtain the second multi-dimensional weight data of the second inspection strategy; the first multi-dimensional weight data includes a first safety weight, a first efficiency weight, and a first cost weight; the second multi-dimensional weight data includes a second safety weight, a second efficiency weight, and a second cost weight;

[0030] The first multidimensional weight data is normalized to obtain the first normalized multidimensional weight data; the second multidimensional weight data is normalized to obtain the second normalized multidimensional weight data.

[0031] The target inspection strategy is obtained by processing the first normalized multidimensional weight data and the second normalized multidimensional weight data using a non-dominated sorting genetic algorithm II.

[0032] A smart inspection device for nuclear power plants includes:

[0033] The first acquisition module is used to acquire real-time monitoring data within the patrol area through the first interface;

[0034] The second acquisition module is used to acquire the real-time inspection records of the inspection robot through the second interface;

[0035] The first strategy module is used to process the real-time monitoring data through the first nuclear power plant inspection strategy evaluation model to obtain the first inspection strategy.

[0036] The second strategy module is used to process the real-time inspection records through the second nuclear power plant inspection strategy evaluation model to obtain the second inspection strategy.

[0037] The target strategy module is used to determine the target inspection strategy based on the first inspection strategy and the second inspection strategy.

[0038] Optionally, the first acquisition module includes:

[0039] A communication unit is established for communicating with multiple nuclear power plant systems within the inspection area via a first interface;

[0040] A raw data acquisition unit is used to acquire raw monitoring data from the plurality of nuclear power plant systems;

[0041] The data filtering unit is used to filter the original monitoring data according to preset filtering rules to obtain the real-time monitoring data.

[0042] Optionally, the second acquisition module includes:

[0043] The raw data acquisition unit is used to collect raw inspection data through the inspection robot.

[0044] The real-time inspection recording unit is used to preprocess the raw inspection data on the robot to obtain the real-time inspection record; the real-time inspection record includes path trajectory, image recognition results and abnormal event records;

[0045] The real-time inspection record acquisition unit is used to acquire the real-time inspection record from the robot terminal in a timed manner and / or an event-triggered manner.

[0046] Optionally, the first nuclear power plant inspection strategy evaluation model includes a nuclear power plant equipment risk prediction model and a nuclear power plant inspection planning model; the nuclear power plant equipment risk prediction model is a hybrid model based on Bayesian networks and long short-term memory networks; the first strategy module includes:

[0047] The model prediction unit is used to process the real-time monitoring data through the nuclear power plant equipment risk prediction model to obtain the predicted status of the nuclear power plant equipment.

[0048] The first strategy generation unit is used to process the predicted status of nuclear power plant equipment through the nuclear power plant inspection planning model to obtain the first inspection strategy.

[0049] Optionally, the second strategy module includes:

[0050] A spatial unit is set up to set the action space and state space of the second nuclear power plant inspection strategy evaluation model according to the real-time inspection records.

[0051] The route planning unit is used to plan multiple candidate inspection paths in the action space;

[0052] The reward value calculation unit is used to calculate the reward value of the candidate inspection path in the state space according to the preset reward function;

[0053] The second inspection path determination unit is used to determine the candidate inspection path with the highest reward value as the second inspection path;

[0054] A second inspection strategy unit is determined, which is used to generate the second inspection strategy based on the second inspection path.

[0055] Optionally, the target strategy module includes:

[0056] A multi-dimensional weight data acquisition unit is used to acquire the first multi-dimensional weight data of the first inspection strategy and the second multi-dimensional weight data of the second inspection strategy. The first multi-dimensional weight data includes a first safety weight, a first efficiency weight, and a first cost weight. The second multi-dimensional weight data includes a second safety weight, a second efficiency weight, and a second cost weight.

[0057] The weight normalization unit is used to normalize the first multidimensional weight data to obtain first normalized multidimensional weight data; and to normalize the second multidimensional weight data to obtain second normalized multidimensional weight data.

[0058] A target inspection strategy unit is defined to process the first normalized multidimensional weight data and the second normalized multidimensional weight data using a non-dominated sorting genetic algorithm II to obtain the target inspection strategy.

[0059] A computer device includes a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, wherein the processor implements the above-described intelligent inspection method for nuclear power plants when executing the computer-readable instructions.

[0060] One or more readable storage media storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the intelligent inspection method for nuclear power plants described above.

[0061] The aforementioned intelligent inspection method, device, computer equipment, and storage medium for nuclear power plants integrate multi-source data from the nuclear power plant monitoring system and inspection robots. It generates inspection strategies emphasizing equipment safety and path optimization using first and second inspection strategy evaluation models, and then comprehensively decides on these strategies to generate a target inspection strategy. This invention achieves dynamic perception of equipment status and intelligent optimization of inspection paths, improving the safety, efficiency, and resource utilization of inspection tasks. This invention embodies multi-model collaboration, data-driven decision-making, and real-time response capabilities, providing strong support for intelligent operation and maintenance of nuclear power plants. This invention can improve the inspection efficiency and quality of nuclear power plants. Attached Figure Description

[0062] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is a flowchart illustrating an intelligent inspection method for nuclear power plants according to one embodiment of the present invention;

[0064] Figure 2 This is a schematic diagram of the structure of an intelligent inspection device for nuclear power plants according to one embodiment of the present invention;

[0065] Figure 3 This is a schematic diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0066] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] In one embodiment, such as Figure 1 As shown, a method for intelligent inspection of a nuclear power plant is provided, including the following steps S10-S50.

[0068] S10. Obtain real-time monitoring data within the patrol area through the first interface.

[0069] Understandably, nuclear power plants have different inspection areas, such as critical areas (reactor buildings), protected areas (nuclear auxiliary buildings), and monitored areas (turbine buildings). Real-time monitoring data within the inspection area can be obtained by connecting to the nuclear power plant control system via the first interface. This real-time monitoring data includes physical parameters such as temperature, pressure, and radiation levels. In one example, real-time monitoring data such as pump body temperature (0-200℃) and cooling water flow rate (0-500 m³ / h) within the inspection area can be collected from the DCS (Digital Control System) via the first interface.

[0070] Optionally, step S10, namely obtaining real-time monitoring data within the patrol area through the first interface, includes:

[0071] S101. Establish communication with multiple nuclear power plant systems within the inspection area through the first interface;

[0072] S102. Obtain raw monitoring data from the multiple nuclear power plant systems;

[0073] S103. Filter the original monitoring data according to the preset filtering rules to obtain the real-time monitoring data.

[0074] Understandably, the first interface refers to the standardized communication interface used to connect to the nuclear power plant control system, supporting multi-protocol compatibility and high-reliability transmission. In some examples, the first interface establishes communication with the DCS system via the Modbus TCP protocol, with the radiation monitoring system (KRS) via the OPC UA protocol, and with the main control room via the WebService protocol.

[0075] Raw monitoring data can be obtained from multiple nuclear power plant systems, such as main steam flow rate (t / h), low-pressure cylinder exhaust temperature (°C), circulating water inlet pressure (MPa), containment hydrogen concentration (%), radiation dose rate (μSv / h), and ventilation system negative pressure gradient. Raw monitoring data is unprocessed data directly collected from nuclear power plant equipment or sensors and may contain redundancy, noise, or outliers.

[0076] Raw monitoring data can be filtered using preset filtering rules to obtain real-time monitoring data. These preset filtering rules are data filtering rules based on nuclear safety guidelines and process requirements, used to extract key and effective information. For example, after Fourier transforming the raw turbine vibration data (sampling rate 10 kHz), non-characteristic frequencies (such as 50 Hz power frequency interference) are filtered out, and the shaft torsional vibration component (0.5-2 kHz) is retained for fault diagnosis.

[0077] This embodiment demonstrates the efficient and accurate acquisition of real-time monitoring data from multiple nuclear power plant systems during a nuclear power plant inspection scenario. First, a stable communication connection is successfully established between the first interface and each key system within the inspection area (S101), ensuring broad coverage and real-time performance of the data source. Next, the raw monitoring data is collected and processed according to preset filtering rules, effectively eliminating irrelevant or redundant information (S102-S103), thereby accurately extracting real-time monitoring data reflecting the status of equipment and the environment. This process not only improves data processing efficiency but also provides high-quality data support for subsequent inspection strategy evaluation.

[0078] S20. Obtain the real-time inspection records of the inspection robot through the second interface.

[0079] Understandably, the second interface is used to interface with various inspection equipment, such as inspection robots. An inspection robot is an autonomous mobile device equipped with multimodal sensors, used for inspection operations in radioactive or high-risk areas of nuclear power plants. Inspection robots include, but are not limited to, quadruped robots, indoor wheeled lifting robots, indoor track-mounted inspection robots, and flying robots. Real-time inspection records from the inspection robot can be obtained through the second interface. These real-time inspection records include, but are not limited to, path trajectories, image recognition results, and abnormal event records. The second interface is specifically designed to interface with various inspection equipment, enhancing the scalability of these devices.

[0080] Optionally, step S20, namely obtaining the real-time inspection records of the inspection robot through the second interface, includes:

[0081] S201. Collect raw inspection data through the inspection robot;

[0082] S202. Preprocess the raw inspection data on the robot to obtain the real-time inspection record; the real-time inspection record includes path trajectory, image recognition results and abnormal event records.

[0083] S203. Obtain the real-time inspection record from the robot terminal according to the timed method and / or the event-triggered method.

[0084] Understandably, inspection robots can be used to collect raw inspection data. Raw inspection data refers to unprocessed data directly output by sensors, such as infrared images, sound waveforms, and gas concentrations. Raw inspection data can be equipment parameters, such as turbine bearing temperature (0-150℃) and main steam pipeline pressure (15-22 MPa), or environmental parameters, such as plant gamma dose rate (0.1-50 μSv / h) and SF6 gas concentration (≤10 ppm).

[0085] The robot terminal can be the inspection robot itself, or it can include the inspection robot itself and a server connected to the inspection robot for preprocessing the raw inspection data. The raw inspection data can be preprocessed on the robot terminal to obtain the real-time inspection record. Preprocessing includes, but is not limited to, data cleaning, feature extraction, and anomaly labeling. The real-time inspection record is a structured data packet containing three core types of information: path trajectory, image recognition results, and anomaly event records. For example, the path trajectory can be generated using magnetic stripe navigation or SLAM algorithms.

[0086] Real-time inspection records can be retrieved from the robot via scheduled or event-triggered methods. In other words, real-time inspection records can be obtained from the inspection robot (or its backend server). Scheduled uploads refer to automatic data uploads at preset intervals (e.g., every 30 minutes). Event-triggered uploads refer to immediate data uploads when a preset threshold anomaly is detected.

[0087] This embodiment achieves efficient data acquisition and intelligent processing of inspection robots in a nuclear power plant setting. By preprocessing the raw data on the robot, data quality is improved and structured real-time inspection records are extracted, including path trajectories, image recognition results, and abnormal event information. Data acquisition is achieved through timed or event-triggered mechanisms, ensuring the timeliness of information and the flexibility of system response. Overall, this embodiment provides accurate and timely data support for subsequent inspection strategy analysis and decision-making.

[0088] S30. The real-time monitoring data is processed through the first nuclear power plant inspection strategy evaluation model to obtain the first inspection strategy.

[0089] Understandably, the inspection strategy evaluation model for the No. 1 nuclear power plant is an environment-equipment state model. By analyzing real-time monitoring data through the No. 1 nuclear power plant inspection strategy evaluation model, the coupled risks of equipment degradation trends and environmental factors are obtained, and then the first inspection strategy is set based on the coupled risks of equipment degradation trends and environmental factors.

[0090] Optionally, the first nuclear power plant inspection strategy evaluation model includes a nuclear power plant equipment risk prediction model and a nuclear power plant inspection planning model; the nuclear power plant equipment risk prediction model is a hybrid model based on Bayesian networks and long short-term memory networks;

[0091] Step S30, namely, processing the real-time monitoring data through the first nuclear power plant inspection strategy evaluation model to obtain the first inspection strategy, includes:

[0092] S301. Process the real-time monitoring data through the nuclear power plant equipment risk prediction model to obtain the predicted status of the nuclear power plant equipment;

[0093] S302. The predicted status of the nuclear power plant equipment is processed through the nuclear power plant inspection planning model to obtain the first inspection strategy.

[0094] Understandably, a nuclear power plant equipment risk prediction model can be a hybrid model based on Bayesian networks and Long Short-Term Memory (LSTM) networks. Bayesian networks explicitly express the causal chains of events such as equipment failures and environmental anomalies through nodes (variables) and edges (dependencies), updating the probability of events by combining real-time monitoring data, supporting reverse reasoning for failures, and can be used for causal reasoning and fault diagnosis. LSTM is a special type of recurrent neural network (RNN), whose core function lies in capturing long-term dependencies and nonlinear dynamic modeling. LSTM processes high-dimensional time-series data (such as vibration waveforms and temperature curves) and extracts key features (such as spectral energy and rate of change). Bayesian networks receive feature data and combine it with domain knowledge (such as equipment failure logic) to construct a causal graph. LSTM predicts future states (such as "vibration will exceed 55 μm in the next 3 hours"), inputting the prediction into the Bayesian network to update the causal probability. The Bayesian network outputs the risk level (such as "Level II risk") and related factors, guiding the LSTM to adjust prediction weights. By processing real-time monitoring data through the nuclear power plant equipment risk prediction model, the predicted status of nuclear power plant equipment can be obtained. The predicted status of nuclear power plant equipment includes, but is not limited to, equipment risk level, environmental associated risks, and remaining working life.

[0095] The nuclear power plant inspection planning model is a decision-making system that integrates multi-objective optimization. It can generate inspection plans by comprehensively considering the predicted status of nuclear power plant equipment, such as equipment risk level, radiation dose, and path efficiency.

[0096] In this embodiment, the first nuclear power plant inspection strategy evaluation model achieves accurate prediction and efficient inspection planning of nuclear power plant equipment and environmental conditions. First, a hybrid model based on Bayesian networks and long short-term memory networks processes real-time monitoring data, accurately identifying equipment degradation trends and environmental coupling risks (S301), providing a scientific basis for inspections. Next, based on these prediction results, the nuclear power plant inspection planning model optimizes and generates a highly targeted and resource-efficient first inspection strategy (S302). This embodiment not only improves the effectiveness and timeliness of inspections but also enhances the safety and stability of nuclear power plant operation.

[0097] S40. The real-time inspection records are processed by the second nuclear power plant inspection strategy evaluation model to obtain the second inspection strategy.

[0098] Understandably, the inspection strategy evaluation model for the second nuclear power plant can employ a reinforcement learning framework (DDPG algorithm). By processing real-time inspection records through the inspection strategy evaluation model for the second nuclear power plant, the second inspection strategy can be obtained.

[0099] Optionally, step S40, namely, processing the real-time inspection records through the second nuclear power plant inspection strategy evaluation model to obtain the second inspection strategy, includes:

[0100] S401. Set the state space and action space of the second nuclear power plant inspection strategy evaluation model according to the real-time inspection records;

[0101] S402. Plan multiple candidate inspection paths in the motion space:

[0102] S403. Calculate the reward value of the candidate inspection path in the state space according to the preset reward function;

[0103] S404. The candidate inspection path with the highest reward value is determined as the second inspection path.

[0104] S405. Generate the second inspection strategy based on the second inspection path.

[0105] Understandably, the state space of the inspection strategy evaluation model for the second nuclear power plant can be updated based on real-time inspection records. The state space represents the context of the current inspection task, including the robot's position, completed paths, uncovered areas, and equipment alarm levels. The action space represents the set of movement strategies that the inspection robot can execute, such as selecting the next inspection target, adjusting the path order, and whether to return to charging.

[0106] In the action space, several candidate inspection paths are generated using search algorithms (such as Monte Carlo tree search, Dijkstra's algorithm, A*, etc.) or reinforcement learning strategies. Each candidate inspection path represents a possible inspection method. For example, path 1: from current location → main pump → control room → return to starting point; path 2: from current location → main pump → cooling tower → charging pile; path 3: from current location → main pump → control room → high temperature alarm point → return.

[0107] A preset reward function is used to evaluate the benefits of each candidate inspection path in the current state space. For example, the preset reward function has several reward items: covering high-risk areas +10 points; checking historical anomalies +5 points; repeating paths -2 points; excessive distance -1 point / meter; failing to approach a charging station when battery is low -5 points. The preset reward function calculates the reward value for each candidate inspection path, and the candidate inspection path with the highest reward value is determined as the second inspection path. The second inspection path is the optimal path selected by the second nuclear power plant inspection strategy evaluation model, used to guide the robot's next stage of inspection.

[0108] Once the second inspection path is determined, a second inspection strategy can be generated based on it. The second inspection strategy can be a specific inspection task instruction, including the path, key inspection points, priority arrangements, etc. In one example, the second inspection strategy = {"Path": ["Main Pump", "Control Room", "High Temperature Alarm Point"],"Priority": ["High Temperature Alarm Point"],"Suggested Inspection Time": "Execute immediately upon next inspection","Remarks": "Observe the main pump's operating status"}.

[0109] This embodiment dynamically constructs the state and action space based on real-time inspection records and optimizes path selection using a reward mechanism, thus achieving intelligent generation of inspection strategies. This method can adaptively adjust inspection paths, prioritizing coverage of high-risk areas and improving the efficiency of abnormal event response. The final output second inspection strategy is highly targeted and real-time, providing a scientific and efficient execution solution for nuclear power plant inspection robots.

[0110] S50. Determine the target inspection strategy based on the first inspection strategy and the second inspection strategy.

[0111] Understandably, the first inspection strategy, based on real-time monitoring data, focuses on equipment health monitoring and environmental risk assessment, characterized by preventative maintenance orientation and high safety. The second inspection strategy, based on real-time inspection records, focuses on path optimization and immediate response, exhibiting high flexibility and adaptability. Combining these two inspection strategies from different perspectives ensures that the target inspection strategy not only meets basic safety requirements but also addresses efficiency and economic considerations in actual operation.

[0112] This embodiment integrates multi-source data from the nuclear power plant monitoring system and inspection robots, and uses first and second inspection strategy evaluation models to generate inspection strategies that emphasize equipment safety and path optimization. Based on these, a comprehensive decision is made to generate the target inspection strategy. This embodiment achieves dynamic perception of equipment status and intelligent optimization of inspection paths, improving the safety, efficiency, and resource utilization of inspection tasks. This embodiment demonstrates multi-model collaboration, data-driven decision-making, and real-time response capabilities, providing strong support for intelligent operation and maintenance of nuclear power plants.

[0113] Optionally, step S50, namely determining the target inspection strategy based on the first inspection strategy and the second inspection strategy, includes:

[0114] S501. Obtain the first multi-dimensional weight data of the first inspection strategy; obtain the second multi-dimensional weight data of the second inspection strategy; the first multi-dimensional weight data includes a first safety weight, a first efficiency weight, and a first cost weight; the second multi-dimensional weight data includes a second safety weight, a second efficiency weight, and a second cost weight.

[0115] S502. Normalize the first multidimensional weight data to obtain first normalized multidimensional weight data; normalize the second multidimensional weight data to obtain second normalized multidimensional weight data.

[0116] S503. The first normalized multidimensional weight data and the second normalized multidimensional weight data are processed by the non-dominated sorting genetic algorithm II to obtain the target inspection strategy.

[0117] Understandably, the first multi-dimensional weight data of the first inspection strategy can be obtained. The first multi-dimensional weight data includes priority weight combinations describing the objective function (safety, efficiency, cost) in the first inspection strategy. The first multi-dimensional weight data includes a first safety weight, a first efficiency weight, and a first cost weight. Taking the intelligent inspection of the pressurized water reactor main pump house as an example, the first safety weight reflects the priority of radiation protection, such as the constraint weight of accumulated radiation dose (e.g., 0.5); the first efficiency weight reflects the timeliness requirement of inspection, such as the optimization weight of path length (e.g., 0.3); and the first cost weight reflects mechanical energy consumption or operation and maintenance costs, such as the weight of the number of turns and angle (e.g., 0.2).

[0118] The second multidimensional weighting data includes priority weight combinations describing the objective functions (safety, efficiency, cost) in the second inspection strategy. The second multidimensional weighting data includes a second safety weight, a second efficiency weight, and a second cost weight.

[0119] The first multidimensional weight data can be normalized to obtain the first normalized multidimensional weight data; the second multidimensional weight data can be normalized to obtain the second normalized multidimensional weight data. By converting weights with different dimensions to a uniform scale (such as the 0-1 interval), the interference of dimensional differences on optimization is eliminated. The weights of the two strategies are normalized separately to ensure that they will not be biased due to differences in numerical range during subsequent fusion.

[0120] The Non-Dominated Sorting Genetic Algorithm II (NSGA-II) can be used to process the first and second normalized multidimensional weight data to obtain the target inspection strategy. NSGA-II is a multi-objective evolutionary algorithm used to find Pareto optimal solutions among multiple conflicting objectives. The NSGA-II algorithm is used to fuse and optimize the normalized weight data of the two strategies, generating a set of Pareto optimal strategies, from which a compromise optimal strategy is selected as the final target inspection strategy.

[0121] This embodiment quantifies and evaluates the weights of the first and second inspection strategies across three key dimensions: safety, efficiency, and cost, and performs normalization to ensure comparability between strategies from different sources. The Non-Dominated Sorting Genetic Algorithm II (NSGA-II) is used to fuse and optimize the multi-objective weights, enabling the search for Pareto optimal solutions among multiple conflicting objectives. The resulting target inspection strategy balances safety with inspection efficiency and cost control, achieving intelligent and comprehensive optimization decision-making for nuclear power plant inspection tasks.

[0122] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0123] In one embodiment, a smart inspection device for nuclear power plants is provided, which corresponds one-to-one with the smart inspection methods for nuclear power plants described in the above embodiments. For example... Figure 2 As shown, the intelligent inspection device for this nuclear power plant includes:

[0124] The first acquisition module 10 is used to acquire real-time monitoring data within the patrol area through the first interface;

[0125] The second acquisition module 20 is used to acquire the real-time inspection records of the inspection robot through the second interface;

[0126] The first strategy module 30 is used to process the real-time monitoring data through the first nuclear power plant inspection strategy evaluation model to obtain the first inspection strategy.

[0127] The second strategy module 40 is used to process the real-time inspection records through the second nuclear power plant inspection strategy evaluation model to obtain the second inspection strategy.

[0128] The target strategy module 50 is used to determine a target inspection strategy based on the first inspection strategy and the second inspection strategy.

[0129] Optionally, the first acquisition module 10 includes:

[0130] A communication unit is established for communicating with multiple nuclear power plant systems within the inspection area via a first interface;

[0131] A raw data acquisition unit is used to acquire raw monitoring data from the plurality of nuclear power plant systems;

[0132] The data filtering unit is used to filter the original monitoring data according to preset filtering rules to obtain the real-time monitoring data.

[0133] Optionally, the second acquisition module 20 includes:

[0134] The raw data acquisition unit is used to collect raw inspection data through the inspection robot.

[0135] The real-time inspection recording unit is used to preprocess the raw inspection data on the robot to obtain the real-time inspection record; the real-time inspection record includes path trajectory, image recognition results and abnormal event records;

[0136] The real-time inspection record acquisition unit is used to acquire the real-time inspection record from the robot terminal in a timed manner and / or an event-triggered manner.

[0137] Optionally, the first nuclear power plant inspection strategy evaluation model includes a nuclear power plant equipment risk prediction model and a nuclear power plant inspection planning model; the nuclear power plant equipment risk prediction model is a hybrid model based on Bayesian networks and long short-term memory networks; the first strategy module 30 includes:

[0138] The model prediction unit is used to process the real-time monitoring data through the nuclear power plant equipment risk prediction model to obtain the predicted status of the nuclear power plant equipment.

[0139] The first strategy generation unit is used to process the predicted status of nuclear power plant equipment through the nuclear power plant inspection planning model to obtain the first inspection strategy.

[0140] Optionally, the second strategy module 40 includes:

[0141] A spatial unit is set up to set the action space and state space of the second nuclear power plant inspection strategy evaluation model according to the real-time inspection records.

[0142] The route planning unit is used to plan multiple candidate inspection paths in the action space;

[0143] The reward value calculation unit is used to calculate the reward value of the candidate inspection path in the state space according to the preset reward function;

[0144] The second inspection path determination unit is used to determine the candidate inspection path with the highest reward value as the second inspection path;

[0145] A second inspection strategy unit is determined, which is used to generate the second inspection strategy based on the second inspection path.

[0146] Optionally, the target strategy module 50 includes:

[0147] A multi-dimensional weight data acquisition unit is used to acquire the first multi-dimensional weight data of the first inspection strategy and the second multi-dimensional weight data of the second inspection strategy. The first multi-dimensional weight data includes a first safety weight, a first efficiency weight, and a first cost weight. The second multi-dimensional weight data includes a second safety weight, a second efficiency weight, and a second cost weight.

[0148] The weight normalization unit is used to normalize the first multidimensional weight data to obtain first normalized multidimensional weight data; and to normalize the second multidimensional weight data to obtain second normalized multidimensional weight data.

[0149] A target inspection strategy unit is defined to process the first normalized multidimensional weight data and the second normalized multidimensional weight data using a non-dominated sorting genetic algorithm II to obtain the target inspection strategy.

[0150] Specific limitations regarding the intelligent inspection device for nuclear power plants can be found in the limitations of the intelligent inspection method for nuclear power plants mentioned above, and will not be repeated here. Each module in the aforementioned intelligent inspection device for nuclear power plants can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0151] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 3As shown, the computer device includes a processor, memory, network interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes a readable storage medium and internal memory. The non-volatile storage medium stores the operating system and computer-readable instructions. The internal memory provides an environment for the operation of the operating system and computer-readable instructions in the readable storage medium. The network interface is used to communicate with an external server via a network connection. When the computer-readable instructions are executed by the processor, they implement a smart inspection method for a nuclear power plant. The readable storage medium provided in this embodiment includes both non-volatile and volatile readable storage media.

[0152] In one embodiment, a computer device is provided, including a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, wherein the processor performs the following steps when executing the computer-readable instructions:

[0153] Real-time monitoring data within the patrol area is obtained through the first interface;

[0154] The real-time inspection records of the inspection robot are obtained through the second interface;

[0155] The real-time monitoring data is processed by the first nuclear power plant inspection strategy evaluation model to obtain the first inspection strategy;

[0156] The real-time inspection records are processed by the second nuclear power plant inspection strategy evaluation model to obtain the second inspection strategy;

[0157] The target inspection strategy is determined based on the first inspection strategy and the second inspection strategy.

[0158] In one embodiment, one or more computer-readable storage media storing computer-readable instructions are provided. The readable storage media provided in this embodiment include non-volatile readable storage media and volatile readable storage media. The readable storage media stores computer-readable instructions, which, when executed by one or more processors, perform the following steps:

[0159] Real-time monitoring data within the patrol area is obtained through the first interface;

[0160] The real-time inspection records of the inspection robot are obtained through the second interface;

[0161] The real-time monitoring data is processed by the first nuclear power plant inspection strategy evaluation model to obtain the first inspection strategy;

[0162] The real-time inspection records are processed by the second nuclear power plant inspection strategy evaluation model to obtain the second inspection strategy;

[0163] The target inspection strategy is determined based on the first inspection strategy and the second inspection strategy.

[0164] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware with computer-readable instructions. These computer-readable instructions can be stored in a non-volatile readable storage medium or a volatile readable storage medium. When executed, these computer-readable instructions can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0165] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0166] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for intelligent inspection of a nuclear power plant, characterized in that, include: Real-time monitoring data within the inspection area is obtained through the first interface; the inspection area includes critical areas, protected areas, and supervised areas in the nuclear power plant; the real-time monitoring data includes temperature, pressure, cooling water flow rate, main steam flow rate, low-pressure cylinder exhaust temperature, hydrogen concentration inside the containment, radiation dose rate, and negative pressure gradient of the ventilation system; The real-time inspection records of the inspection robot are obtained through the second interface; the inspection robot is an autonomous mobile device equipped with multimodal sensors, used for inspection operations in radioactive or high-risk areas of nuclear power plants. The real-time inspection record includes path trajectory, image recognition results, and abnormal event records; the inspection robot includes quadruped robot, indoor wheeled lifting robot, indoor track-mounted inspection robot, and flying robot. The real-time monitoring data is processed by the first nuclear power plant inspection strategy evaluation model to obtain the first inspection strategy; The first inspection strategy focuses on equipment health monitoring and environmental risk assessment; The real-time inspection records are processed by the second nuclear power plant inspection strategy evaluation model to obtain the second inspection strategy; The second inspection strategy focuses on path optimization and immediate response; Determine the target inspection strategy based on the first inspection strategy and the second inspection strategy; The first nuclear power plant inspection strategy evaluation model includes a nuclear power plant equipment risk prediction model and a nuclear power plant inspection planning model; the nuclear power plant equipment risk prediction model is a hybrid model based on Bayesian networks and long short-term memory networks. The process of processing the real-time monitoring data through the first nuclear power plant inspection strategy evaluation model to obtain the first inspection strategy includes: The real-time monitoring data is processed by the nuclear power plant equipment risk prediction model to obtain the predicted status of the nuclear power plant equipment. The nuclear power plant inspection planning model is used to process the predicted status of the nuclear power plant equipment to obtain the first inspection strategy. The process of processing the real-time inspection records using the second nuclear power plant inspection strategy evaluation model to obtain the second inspection strategy includes: The action space and state space of the second nuclear power plant inspection strategy evaluation model are set according to the real-time inspection records. Multiple candidate inspection paths are planned within the aforementioned action space; The reward value of the candidate inspection path in the state space is calculated according to the preset reward function; the preset reward function has multiple reward items: covering high-risk areas +10 points; checking historical anomalies +5 points; repeating the path -2 points; distance too long -1 point / meter; not approaching the charging pile when the battery is low -5 points; The inspection path with the highest reward value is selected as the second inspection path. The second inspection strategy is generated based on the second inspection path; The step of determining the target inspection strategy based on the first inspection strategy and the second inspection strategy includes: Obtain the first multidimensional weight data of the first inspection strategy; obtain the second multidimensional weight data of the second inspection strategy; the first multidimensional weight data includes a first safety weight, a first efficiency weight, and a first cost weight; the second multidimensional weight data includes a second safety weight, a second efficiency weight, and a second cost weight; The first multidimensional weight data is normalized to obtain the first normalized multidimensional weight data; the second multidimensional weight data is normalized to obtain the second normalized multidimensional weight data. The target inspection strategy is obtained by processing the first normalized multidimensional weight data and the second normalized multidimensional weight data using a non-dominated sorting genetic algorithm II.

2. The intelligent inspection method for nuclear power plants as described in claim 1, characterized in that, The process of acquiring real-time monitoring data within the patrol area through the first interface includes: Communication is established with multiple nuclear power plant systems within the inspection area through the first interface; Raw monitoring data were obtained from the multiple nuclear power plant systems; The original monitoring data is filtered according to preset filtering rules to obtain the real-time monitoring data.

3. The intelligent inspection method for nuclear power plants as described in claim 1, characterized in that, The step of obtaining the real-time inspection records of the inspection robot through the second interface includes: The inspection robot collects raw inspection data. The robot preprocesses the raw inspection data to obtain the real-time inspection record; The real-time inspection records are obtained from the robot terminal using a timed method and / or an event-triggered method.

4. A smart inspection device for nuclear power plants, characterized in that, include: The first acquisition module is used to acquire real-time monitoring data within the inspection area through the first interface; the inspection area includes the critical area, protection area, and supervision area in the nuclear power plant; the real-time monitoring data includes temperature, pressure, cooling water flow rate, main steam flow rate, low-pressure cylinder exhaust temperature, hydrogen concentration inside the containment, radiation dose rate, and negative pressure gradient of the ventilation system; The second acquisition module is used to acquire the real-time inspection records of the inspection robot through the second interface; the inspection robot is an autonomous mobile device equipped with multimodal sensors, used for inspection operations in radioactive or high-risk areas of nuclear power plants; the real-time inspection records include path trajectory, image recognition results, and abnormal event records; the inspection robot includes quadruped robots, indoor wheeled lifting robots, indoor track-type inspection robots, and flying robots; The first strategy module is used to process the real-time monitoring data through the first nuclear power plant inspection strategy evaluation model to obtain the first inspection strategy. The first inspection strategy focuses on equipment health monitoring and environmental risk assessment; The second strategy module is used to process the real-time inspection records through the second nuclear power plant inspection strategy evaluation model to obtain the second inspection strategy. The second inspection strategy focuses on path optimization and immediate response; The target strategy module is used to determine the target inspection strategy based on the first inspection strategy and the second inspection strategy; The first nuclear power plant inspection strategy evaluation model includes a nuclear power plant equipment risk prediction model and a nuclear power plant inspection planning model; the nuclear power plant equipment risk prediction model is a hybrid model based on Bayesian networks and long short-term memory networks; the first strategy module includes: The model prediction unit is used to process the real-time monitoring data through the nuclear power plant equipment risk prediction model to obtain the predicted status of the nuclear power plant equipment. The first strategy generation unit is used to process the predicted status of nuclear power plant equipment through the nuclear power plant inspection planning model to obtain the first inspection strategy. The second strategy module includes: A spatial unit is set up to set the action space and state space of the second nuclear power plant inspection strategy evaluation model according to the real-time inspection records. The route planning unit is used to plan multiple candidate inspection paths in the action space; The reward value calculation unit is used to calculate the reward value of the candidate inspection path in the state space according to the preset reward function; the preset reward function is set with multiple reward items: covering high-risk areas +10 points; checking historical anomalies +5 points; repeating paths -2 points; distance too long -1 point / meter; not approaching the charging pile when the battery is low -5 points; The second inspection path determination unit is used to determine the candidate inspection path with the highest reward value as the second inspection path; A second inspection strategy unit is determined, which is used to generate the second inspection strategy based on the second inspection path; The target strategy module includes: A multi-dimensional weight data acquisition unit is used to acquire the first multi-dimensional weight data of the first inspection strategy and the second multi-dimensional weight data of the second inspection strategy. The first multi-dimensional weight data includes a first safety weight, a first efficiency weight, and a first cost weight. The second multi-dimensional weight data includes a second safety weight, a second efficiency weight, and a second cost weight. The weight normalization unit is used to normalize the first multidimensional weight data to obtain first normalized multidimensional weight data; and to normalize the second multidimensional weight data to obtain second normalized multidimensional weight data. A target inspection strategy unit is defined to process the first normalized multidimensional weight data and the second normalized multidimensional weight data using a non-dominated sorting genetic algorithm II to obtain the target inspection strategy.

5. The intelligent inspection device for nuclear power plants as described in claim 4, characterized in that, The first acquisition module includes: A communication unit is established for communicating with multiple nuclear power plant systems within the inspection area via a first interface; A raw data acquisition unit is used to acquire raw monitoring data from the plurality of nuclear power plant systems; The data filtering unit is used to filter the original monitoring data according to preset filtering rules to obtain the real-time monitoring data.

6. The intelligent inspection device for nuclear power plants as described in claim 4, characterized in that, The second acquisition module includes: The raw data acquisition unit is used to collect raw inspection data through the inspection robot. The real-time inspection recording unit is used to preprocess the raw inspection data on the robot end to obtain the real-time inspection record; The real-time inspection record acquisition unit is used to acquire the real-time inspection record from the robot terminal in a timed manner and / or an event-triggered manner.

7. A computer device comprising a memory, a processor, and computer-readable instructions stored in the memory and running on the processor, characterized in that, When the processor executes the computer-readable instructions, it implements the intelligent inspection method for nuclear power plants as described in any one of claims 1 to 3.

8. One or more readable storage media storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the intelligent inspection method for a nuclear power plant as described in any one of claims 1 to 3.

Images