A power plant intelligent inspection system based on large model

Abstract

The present invention relates to the technical field of power plant production inspection. The present invention provides a power plant intelligent inspection system based on a large model. The present invention realizes data transmission and processing through communication connection between the server side and multiple inspection devices and the inspected equipment. The present invention includes multiple units such as sensor fault detection, drone inspection optimization, data fusion analysis, and data acquisition fault repair, aiming to solve the inspection distortion problem caused by sensor failure or data transmission interference, and at the same time improve the inspection stability and accuracy of drones under harsh conditions. The present invention uses a random forest algorithm to construct a sensor data model, dynamically adjusts the drone inspection path, and performs data fusion analysis through deep learning to generate accurate inspection conclusions and disposal plans. The present invention not only improves the accuracy and efficiency of inspections, but also further enhances the intelligence and adaptability of the system.

Description

Technical Field

[0001] The present invention relates to the technical field of power plant production inspection, and in particular to a power plant intelligent inspection system based on a large model. Background Art

[0002] In the field of power plant production inspection, the initial manual inspection has been gradually replaced by intelligent technology. At present, intelligent inspection systems and drone inspection technologies have become the mainstream in the industry. The intelligent inspection system integrates cutting-edge technologies such as IoT sensors, edge computing, and artificial intelligence, which promotes the digitalization and intelligentization of power plant inspection work. By introducing these advanced technologies, power plants can monitor the operating status of equipment more efficiently and accurately.

[0003] The application of intelligent inspection systems enables power plants to monitor various parameters of equipment in real time, such as temperature, pressure, vibration, etc., and detect abnormal conditions in a timely manner. The system performs preliminary data processing through edge computing, and then uses artificial intelligence technology for in-depth analysis, thereby realizing intelligent assessment of the health status of equipment. Drone inspection technology further broadens the field of vision and scope of inspection. By carrying high-definition cameras and sensors, drones can capture the status of equipment in areas that are difficult to reach manually, and combine data processing and software technology to provide power plants with comprehensive equipment operation reports.

[0004] Although the intelligent inspection system and drone technology have significantly improved the efficiency and accuracy of power plant equipment inspections, there are still the following technical pain points. The intelligent inspection system relies on a large number of sensors and data processing capabilities. When some sensors fail or data transmission is disturbed, the inspection results are distorted. Although drone inspections have a wide coverage, their stability and accuracy will be affected in complex environments, such as strong winds, low visibility and other severe weather conditions, resulting in reduced accuracy of the intelligent inspection system, which in turn affects the generation of accurate inspection conclusions and disposal plans by the big data model. Summary of the invention

[0005] The purpose of the present invention is to provide a power plant intelligent inspection system based on a large model to solve the problem of distortion of inspection results when certain sensors fail or data transmission is disturbed. Although drone inspections have a wide coverage, their stability and accuracy will be affected in complex environments, such as strong winds, low visibility and other severe weather conditions, resulting in reduced accuracy of the intelligent inspection system, thereby affecting the generation of accurate inspection conclusions and disposal plans by the big data model.

[0006] The present invention provides a power plant intelligent inspection system based on a large model, including a server end, a background control end, an inspection device end, an inspected device end and a user end. The server end establishes a communication connection with the background control end, the inspection device end, the inspected device end and the user end. The server end is used to process and analyze data transmitted by the background control end, the inspection device end, the inspected device end and the user end. The server end includes:

[0007] The sensor fault detection unit obtains the data collected by the sensors on the intelligent inspection equipment and the power equipment, uses the random forest algorithm to train the data collected by the sensors, builds a sensor data model, predicts the sensor data through the sensor data model, obtains the model prediction data, compares the model prediction data with the real-time data, and identifies the data points that deviate from the normal range by calculating the deviation between the real-time data and the model prediction data. When multiple consecutive data points are marked as abnormal, the sensor fault warning information is issued. If the data points that deviate from the normal range are identified to be within the preset safety range, the data collected by the sensor are marked as usable detection data;

[0008] The drone inspection optimization unit compares the available detection data with the preset safety values ​​of each device. If there is no abnormal data in the comparison result, the drone inspection will be carried out according to the preset path. If there is abnormal data in the comparison result, the location information corresponding to the abnormal data will be obtained and substituted into the drone inspection path generation model to generate the drone collaborative inspection path.

[0009] The data fusion analysis unit receives and processes data from multiple sensors and drones for data fusion, uses deep learning algorithms to perform in-depth analysis on the fused data to identify potential faults, builds a decision support model, substitutes the identified potential faults into the decision support model, and generates information on fault location and maintenance suggestions for maintenance personnel;

[0010] The data collection fault repair unit generates inspection conclusions through a pre-trained large model based on the fault location and repair suggestions provided by the maintenance personnel. The inspection conclusions include equipment status reports, abnormality lists and risk levels, and automatically generate inspection disposal plans;

[0011] The re-inspection path generation unit generates the inspection path for the drone re-inspection based on the location information corresponding to the recorded abnormal data and the specific location of the fault repair. The data collected by the drone re-inspection is transmitted to the data fusion analysis unit.

[0012] Furthermore, the present invention provides a large-model-based intelligent inspection system for power plants, a sensor fault detection unit, and further includes:

[0013] Obtain the operation and management data of the power plant, including historical inspection records, equipment status, fault reports, environmental monitoring data, etc., clean the data to remove errors and outliers, and annotate the data to distinguish the normal, warning and fault status of the equipment;

[0014] Data features are extracted from the operation and maintenance and management data of the power plant, and the extracted data features are divided into training set, validation set and test set. The data features in the training set, validation set and test set are divided into time windows, and anomaly detection is performed to identify outliers in the data.

[0015] Furthermore, the present invention provides a large-model-based power plant intelligent inspection system, a data fusion analysis unit, comprising:

[0016] Train the big model based on the data generated in the sensor fault detection unit and define the architecture of the big model, including the configuration of the attention mechanism, the number of encoder and decoder layers, and the number of heads;

[0017] The preprocessed data set generated in the sensor fault detection unit is fed into the large model, and the self-attention mechanism is used to capture the long-range dependencies in the data;

[0018] Through the optimizer and learning rate scheduling strategy, the training process of the large model is started to obtain the trained large model.

[0019] Furthermore, the present invention provides a large-model-based intelligent inspection system for power plants, and an unmanned aerial vehicle inspection optimization unit, comprising:

[0020] Input the power plant's layout information, equipment distribution, historical fault data, and inspection record data into the trained large model for analysis;

[0021] The trained large model combines machine learning and optimization algorithms to analyze the criticality, failure rate, interdependence and periodic inspection requirements of equipment, and generates one or more optimized inspection paths.

[0022] Furthermore, the present invention provides a large-model-based intelligent inspection system for power plants, a sensor fault detection unit, comprising:

[0023] Use the trained big model to analyze the data collected by the drone to automatically identify abnormalities and potential safety risks of the inspected equipment. The trained big model conducts in-depth analysis of the data and provides fault diagnosis and risk assessment for the inspected equipment.

[0024] The image recognition algorithm is used to detect abnormal conditions on the inspected equipment, including cracks, corrosion, and foreign body intrusion. The thermal imaging analysis algorithm is used to identify potential faults such as overheating and insulation damage of the inspected equipment. Combined with the historical inspection data and real-time monitoring results, the trained large model is used to predict the failure trend of the inspected equipment.

[0025] Furthermore, the present invention provides a large-model-based intelligent inspection system for power plants, a data acquisition fault repair unit, and further includes:

[0026] The inspection and disposal plan is used to guide power maintenance personnel in carrying out maintenance and repair work. The inspection and disposal plan includes inspection problem description, location positioning, severity assessment and fault cause analysis.

[0027] Furthermore, the present invention provides a large-model-based intelligent inspection system for power plants, a data acquisition fault repair unit, and further includes:

[0028] Match the exception list and risk level in the historical database and knowledge base to obtain the cases and solutions corresponding to the exception list and risk level;

[0029] Based on the current equipment status, available resources and the maintenance personnel's successful maintenance cases, the maintenance plan is matched to generate an inspection and disposal plan for the current fault. The inspection and disposal plan includes emergency repair steps, preventive maintenance suggestions and parts that need to be replaced.

[0030] Furthermore, the present invention provides a large-model-based intelligent inspection system for power plants, a data acquisition fault repair unit, and further includes:

[0031] Conduct preliminary screening of received exception lists and risk level information to remove duplicate, invalid or obviously erroneous data;

[0032] Identify and determine abnormal data by comparing it with the normal parameter range of the abnormal list;

[0033] According to the source of abnormal information, combined with the equipment layout and monitoring system data of the power plant, the specific location of the abnormal data is located.

[0034] Furthermore, the present invention provides a large-model-based intelligent inspection system for power plants, a data acquisition fault repair unit, and further includes:

[0035] Use abnormal information and risk level as search keywords to search in the historical database and find solutions and cases related to the current fault in the knowledge base;

[0036] According to the similar fault cases in the historical records and the information in the knowledge base, the solutions to the abnormal information are matched to obtain the approximate solutions to the abnormal information.

[0037] Furthermore, the present invention provides a large-model-based intelligent inspection system for power plants, a data acquisition fault repair unit, and further includes:

[0038] Based on the approximate solution of abnormal information, the data format is converted to obtain the inspection and disposal plan.

[0039] The inspection and treatment plan includes emergency repair steps, preventive maintenance suggestions, and information on parts that need to be replaced. The generated inspection and treatment plan is fed back to the maintenance personnel.

[0040] The beneficial effects of the present invention are as follows:

[0041] Improve inspection accuracy: Through the sensor fault detection unit, the present invention can promptly detect and handle sensor faults or data transmission interference, ensuring that only accurate data is included in the analysis, thereby significantly improving the accuracy of inspection results.

[0042] Improve the stability and efficiency of drone inspections: The drone inspection optimization unit enables drones to flexibly adjust inspection routes under severe weather conditions and avoid adverse environmental factors, which not only ensures the stability of inspections but also improves inspection efficiency.

[0043] Data fusion improves decision reliability: The data fusion analysis unit fuses data from different sensors and performs in-depth analysis through deep learning algorithms, providing more reliable data support for fault location and maintenance recommendations, thereby improving the accuracy of decision-making.

[0044] Quick response and repair: The data collection fault repair unit can quickly generate inspection conclusions and disposal plans based on fault location and maintenance suggestions, shortening fault response time and improving repair efficiency.

[0045] Improve the intelligence level of the system: Through the training and application of big data models, the present invention achieves a higher degree of intelligence, can automatically identify potential problems and failures, and reduce the need for human intervention.

[0046] Enhanced system robustness: The comprehensive application of the present invention enables the intelligent inspection system to maintain efficient operation in the face of various complex environments and situations, significantly enhancing the robustness of the system.

[0047] In summary, the present invention provides strong technical support for intelligent inspection in complex industrial environments such as power plants by improving inspection accuracy, improving the stability and efficiency of drone inspections, improving decision-making reliability through data fusion, quickly responding to and repairing faults, and improving the intelligence and robustness of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0049] Figure 1 A schematic diagram of a system architecture provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. The technical solutions provided by the embodiments of the present invention are described in detail below in conjunction with the drawings.

[0051] See also Figure 1 The present invention provides a power plant intelligent inspection system based on a large model, including a server end, a background control end, an inspection device end, an inspected device end and a user end. The server end establishes a communication connection with the background control end, the inspection device end, the inspected device end and the user end. The server end is used to process and analyze data transmitted by the background control end, the inspection device end, the inspected device end and the user end. The server end includes:

[0052] The sensor fault detection unit obtains the data collected by the sensors on the intelligent inspection equipment and the power equipment, uses the random forest algorithm to train the data collected by the sensors, builds a sensor data model, predicts the sensor data through the sensor data model, obtains the model prediction data, compares the model prediction data with the real-time data, and identifies the data points that deviate from the normal range by calculating the deviation or abnormality between the real-time data and the model prediction data. When multiple consecutive data points are marked as abnormal, the sensor fault warning information is issued. If the data points that deviate from the normal range are identified to be within the preset safety range, the data collected by the sensor are marked as usable detection data;

[0053] The server first obtains the data collected by the sensors on the intelligent inspection equipment and power equipment through the sensor fault detection unit. Then, the random forest algorithm is used to train these data to build a sensor data model. The sensor data model can predict the normal data of the sensor. The real-time collected data will be compared with the data predicted by this model. By calculating the deviation or abnormality between the two, the system can identify those data points that deviate from the normal range. If multiple consecutive data points are marked as abnormal, the system will issue a sensor fault warning message. If the identified abnormal data is still within the safe range, this data will be marked as usable detection data.

[0054] Function and effect of sensor fault detection unit: The main function of sensor fault detection unit is to ensure the accuracy and reliability of sensor data. By detecting sensor data in real time, sensor failure or problems in data transmission can be discovered in time, thus avoiding the impact of erroneous data on subsequent analysis. This improves the stability and accuracy of the entire inspection system.

[0055] The drone inspection optimization unit compares the available detection data with the preset safety values ​​of each device. If there is no abnormal data in the comparison result, the drone inspection will be carried out according to the preset path. If there is abnormal data in the comparison result, the location information corresponding to the abnormal data will be obtained and substituted into the drone inspection path generation model to generate the drone collaborative inspection path.

[0056] In the drone inspection optimization unit, the server will compare the inspection data previously marked as usable with the preset equipment safety value. If no abnormal data is found in the comparison result, the drone will inspect according to the preset path. However, if there is abnormal data in the comparison result, the server will obtain the location information corresponding to these abnormal data and input this information into the drone inspection path generation model to generate an optimized and coordinated inspection path.

[0057] The role and effect of the drone inspection optimization unit: The role of the drone inspection optimization unit is to dynamically adjust the inspection path of the drone based on real-time data. This not only improves the efficiency of the inspection, but also ensures that the drone can quickly reach areas where there may be problems for detailed inspection. This dynamically adjusted path planning method enhances the flexibility and pertinence of drone inspections.

[0058] The data fusion analysis unit receives and processes data from multiple sensors and drones for data fusion, uses deep learning algorithms to perform in-depth analysis on the fused data to identify potential faults, builds a decision support model, substitutes the identified potential faults into the decision support model, and generates information on fault location and maintenance suggestions for maintenance personnel;

[0059] The server receives and processes data from multiple sensors and drones through a data fusion analysis unit to perform data fusion. Then, it uses a deep learning algorithm to perform in-depth analysis on the fused data to identify potential faults. In addition, the server will also substitute these identified faults into a pre-built decision support model to provide maintenance personnel with accurate fault location and maintenance suggestions.

[0060] The present invention takes the following steps in constructing a decision support model:

[0061] Data collection and preprocessing: Collect inspection data from various sensors and drones. Clean and preprocess the collected data to remove noise and outliers to ensure data quality.

[0062] Data fusion: Use data fusion technology to combine data from different sensors to form a comprehensive data set.

[0063] Deep learning analysis: Use deep learning algorithms to conduct in-depth analysis on the fused data to identify potential failure modes and risks.

[0064] Model definition and training: Define the architecture of the decision support model including the configuration of the attention mechanism, the number of encoder and decoder layers, the number of heads, etc. Input the preprocessed dataset into the model for training, and optimize the model performance through the optimizer and learning rate scheduling strategy.

[0065] Model integration and application: Integrate the trained deep learning model into the decision support system. Substitute the identified potential faults into the decision support model to generate more accurate fault location and maintenance recommendation information.

[0066] Decision support: Use decision support models to provide maintenance personnel with accurate fault location, maintenance suggestions, and fault cause analysis.

[0067] Continuous optimization: Continuously optimize the decision support model based on actual application conditions to improve its accuracy and reliability.

[0068] Through these steps, the present invention constructs a decision support model based on deep learning, which can integrate multi-source inspection data, provide reliable fault location and maintenance suggestions, and thus support the intelligent inspection and maintenance of power plant equipment. The specific technical implementation includes the use of advanced deep learning architectures such as the Transformer model, as well as customized model development and optimization based on the specific operation and maintenance data and business needs of the power plant.

[0069] Function and effect of data fusion analysis unit: The main function of data fusion analysis unit is to provide more accurate and comprehensive fault diagnosis through data fusion and deep analysis. This multi-source data fusion method can reduce the error and uncertainty caused by a single data source, thereby improving the accuracy and reliability of fault detection. At the same time, through the analysis of deep learning algorithms, it can more deeply explore potential problems in the data and provide maintenance personnel with more valuable maintenance suggestions.

[0070] The data collection fault repair unit generates inspection conclusions through a pre-trained large model based on the fault location and repair suggestions provided by the maintenance personnel. The inspection conclusions include equipment status reports, abnormality lists and risk levels, and automatically generate inspection disposal plans;

[0071] In the data collection and fault repair unit, the server will generate inspection conclusions through a pre-trained large model based on the fault location and repair suggestions provided by the maintenance personnel. This conclusion includes equipment status reports, abnormality lists and risk levels, and will automatically generate an inspection disposal plan.

[0072] Function and effect of data acquisition fault repair unit: The function of data acquisition fault repair unit is to quickly respond to and handle detected faults and problems. By automatically generating inspection and disposal plans, it can shorten the time for fault response and processing and improve maintenance efficiency. At the same time, the equipment status report, abnormality list and risk level information provided can also help maintenance personnel understand the operating status of the equipment more comprehensively, so as to make more accurate maintenance decisions.

[0073] The re-inspection path generation unit regenerates the inspection path for the drone re-inspection based on the location information corresponding to the recorded abnormal data and the specific location of the fault repair. The data collected by the drone re-inspection is transmitted back to the data fusion analysis unit.

[0074] The server will regenerate a drone inspection route based on the location information corresponding to the previously recorded abnormal data and the specific location of the fault repair. The drone will re-inspect along this newly generated route and transmit the collected data back to the data fusion analysis unit for further analysis and processing.

[0075] Function and effect of the re-inspection path generation unit: The function of the re-inspection path generation unit is to ensure that the area after the fault is repaired can be re-inspected and verified in a timely manner. By re-planning the inspection path of the drone, it can ensure that the drone can quickly reach and inspect the area with previous problems in detail, thereby verifying the repair effect and ensuring the normal operation of the equipment. This re-inspection mechanism improves the reliability and integrity of the entire inspection system.

[0076] Specifically, the present invention provides a large-model-based intelligent inspection system for power plants, a sensor fault detection unit, and further includes:

[0077] Obtain the operation and management data of the power plant, including historical inspection records, equipment status, fault reports, environmental monitoring data, etc., clean the data to remove errors and outliers, and annotate the data to distinguish the normal, warning and fault status of the equipment;

[0078] Data features are extracted from the operation and maintenance and management data of the power plant, and the extracted data features are divided into training set, validation set and test set. The data features in the training set, validation set and test set are divided into time windows, and anomaly detection is performed to identify abnormal values ​​or patterns in the data.

[0079] Specifically, the present invention provides a large-model-based power plant intelligent inspection system, a data fusion analysis unit, comprising:

[0080] Train the big model based on the data generated in the sensor fault detection unit and define the architecture of the big model, including the configuration of the attention mechanism, the number of encoder and decoder layers, and the number of heads;

[0081] The cleaned and preprocessed data generated in the sensor fault detection unit is input into the large model for encoding representation, and the Transformer's self-attention mechanism is used to capture long-distance dependencies in the data;

[0082] Through the optimizer and learning rate scheduling strategy, the training process of the large model is started to obtain the trained large model.

[0083] Specifically, the present invention provides a large-model-based power plant intelligent inspection system and a drone inspection optimization unit, comprising:

[0084] Input the power plant's layout information, equipment distribution, historical fault data, and inspection record data into the trained large model for analysis;

[0085] The trained large model combines machine learning and optimization algorithms to analyze the criticality, failure rate, interdependence and periodic inspection requirements of equipment, and generates one or more optimized inspection paths.

[0086] Specifically, the present invention provides a large-model-based power plant intelligent inspection system, a sensor fault detection unit, comprising:

[0087] Use the trained big model to analyze the data collected by the drone to automatically identify abnormalities and potential safety risks of the inspected equipment. The trained big model conducts in-depth analysis of the data to provide fault diagnosis and risk assessment of the inspected equipment;

[0088] The image recognition algorithm is used to detect abnormal conditions on the inspected equipment, including cracks, corrosion, and foreign body intrusion. The thermal imaging analysis algorithm is used to identify potential faults such as overheating and insulation damage of the inspected equipment. Combined with the historical inspection data and real-time monitoring results, the trained large model is used to predict the failure trend of the inspected equipment.

[0089] Specifically, the present invention provides a large-model-based power plant intelligent inspection system, a data acquisition fault repair unit, and further includes:

[0090] The inspection and disposal plan is used to guide power maintenance personnel to carry out necessary maintenance and repair work. The inspection and disposal plan includes inspection problem description, location positioning, severity assessment and fault cause analysis.

[0091] Specifically, the present invention provides a large-model-based power plant intelligent inspection system, a data acquisition fault repair unit, and further includes:

[0092] According to the matching of the exception list and risk level in the historical database and knowledge base, the cases and solutions corresponding to the exception list and risk level are obtained. The maintenance plan is matched based on the current equipment status, available resources and the maintenance personnel's successful maintenance cases, and the inspection and disposal plan for the current fault is generated. The inspection and disposal plan includes emergency repair steps, preventive maintenance suggestions and parts that need to be replaced.

[0093] Specifically, the present invention provides a large-model-based power plant intelligent inspection system, a data acquisition fault repair unit, and further includes:

[0094] Conduct preliminary screening of received exception lists and risk level information to remove duplicate, invalid or obviously erroneous data;

[0095] Identify and determine abnormal data by comparing it with the normal parameter range of the abnormal list;

[0096] According to the source of abnormal information, combined with the equipment layout and monitoring system data of the power plant, locate the specific location of abnormal data;

[0097] Specifically, the present invention provides a large-model-based power plant intelligent inspection system, a data acquisition fault repair unit, and further includes:

[0098] Use abnormal information and risk level as search keywords to search in the historical database and find solutions and cases related to the current fault in the knowledge base;

[0099] According to the similar fault cases in the historical records and the information in the knowledge base, the solutions to the abnormal information are matched to obtain the approximate solutions to the abnormal information.

[0100] Specifically, the present invention provides a large-model-based power plant intelligent inspection system, a data acquisition fault repair unit, and further includes:

[0101] Based on the approximate solution of abnormal information, the data format is converted to obtain the inspection and disposal plan.

[0102] The inspection and disposal plan includes emergency repair steps, preventive maintenance suggestions, and information on parts that need to be replaced. The generated inspection and disposal plan is fed back to maintenance personnel.

[0103] In the data acquisition and fault repair unit of the intelligent inspection system of the power plant, a number of advanced technologies are used to ensure the stable operation of the power plant equipment. First, through the application of large models, the data acquisition and fault repair unit can intelligently generate inspection conclusions, including equipment status reports, abnormality lists and risk levels, providing strong support for subsequent analysis and decision-making. Secondly, by matching the historical database and knowledge base, the system can quickly find historical cases and solutions similar to the current fault, providing valuable references for maintenance personnel. In addition, the maintenance plan matching technology recommends the most suitable maintenance plan for maintenance personnel based on the current equipment status, resources and successful cases. At the same time, data screening and positioning technology helps the system to quickly and accurately identify and locate abnormal data, improving maintenance efficiency. Finally, through the solution generation and feedback mechanism, the system can generate specific inspection and disposal plans based on historical data and knowledge bases, and provide timely feedback to maintenance personnel. Throughout the process, the application of automation and intelligent technology reduces manual intervention and improves the speed of fault response and repair.

[0104] In the data preprocessing stage, the received abnormal list and risk level information are initially screened to remove duplicate, invalid or obviously erroneous data to ensure the accuracy of subsequent analysis. At the same time, by comparing abnormal data with the normal parameter range, the system can identify specific problem points.

[0105] Secondly, in the fault location and assessment phase, combined with the equipment layout and monitoring system data of the power plant, the server can accurately locate the specific location of the abnormality and evaluate the severity of the abnormality, providing an important basis for subsequent repair work. Through historical case matching, the server uses abnormal information and risk levels to search for similar cases in the historical database and find relevant solutions from the knowledge base, so as to quickly find an approximate solution to the current fault.

[0106] Then, in the solution generation stage, the server will convert the data format based on the approximate solution found, and generate an inspection and disposal plan suitable for the current fault. The plan lists in detail the emergency repair steps, preventive maintenance suggestions, and information on parts that need to be replaced, ensuring that maintenance personnel can carry out repair work quickly and accurately.

[0107] In addition, the server can automatically generate maintenance suggestions and disposal plans for specific faults through pre-built decision support models. At the same time, when generating inspection and disposal plans, the system also considers the current equipment status, available resources and maintenance personnel's successful cases.

[0108] Finally, in the feedback and closed-loop stage, the server will promptly feed back the generated inspection and disposal plan to the maintenance personnel so that they can quickly respond and repair the fault. At the same time, the system also has a re-inspection mechanism to re-plan the inspection route of the drone after the fault is repaired to ensure that the repair effect is verified.

[0109] The present invention uses multiple units, such as sensor fault detection, UAV inspection optimization, data fusion analysis, and data acquisition fault repair, combined with in-depth analysis of big data models, to effectively solve the problem of distorted inspection results caused by sensor failure or data transmission interference, while improving the stability and accuracy of UAV inspections under severe weather conditions.

[0110] First, the sensor fault detection unit can detect and warn of sensor failures in a timely manner, ensuring that only accurate data is used for subsequent analysis. The drone inspection optimization unit dynamically adjusts the inspection path according to abnormal conditions, avoiding harsh environments and improving inspection efficiency.

[0111] The data fusion and analysis unit deeply fuses and analyzes data from multiple sensors and drones, improving the accuracy and reliability of inspection data. The big data model plays a key role in this process. The big data model uses deep learning algorithms to identify potential problems and faults, and generates fault location and maintenance recommendations through decision support models.

[0112] Finally, the data acquisition fault repair unit generates inspection conclusions including equipment status reports, abnormality lists, risk levels, and specific disposal plans based on these fault locations and repair suggestions. These plans include emergency repair steps, preventive maintenance suggestions, and parts that need to be replaced, thus ensuring the accuracy of the inspection conclusions and the effectiveness of the disposal plans.

[0113] In summary, the present invention comprehensively improves the accuracy and reliability of the intelligent inspection system through the collaborative work of multiple units and in-depth analysis of big data models, and provides an effective solution to the inspection problems of sensor failure, data transmission interference and severe weather conditions.

[0114] The implementation method of the power plant intelligent inspection system based on the large model provided by the present invention involves multiple steps, which are specifically as follows:

[0115] First, sensor fault detection is performed to collect sensor data. The data collected by the sensor is collected in real time through the communication connection between the server and the sensor device. The collected sensor data is modeled using the random forest algorithm or other machine learning algorithms to identify the normal data range. The real-time sensor data is compared with the normal range predicted by the model to identify and mark data points that deviate from the normal range, thereby detecting sensor faults or data transmission interference.

[0116] According to the abnormal data marked by the sensor fault detection unit, the approximate location of the abnormality is determined. According to the location information of the abnormal data, the inspection path of the drone is dynamically adjusted to ensure that the drone can efficiently and accurately cover the abnormal area, while avoiding harsh environments such as strong winds and low visibility. The drone conducts inspections along the planned path and collects data such as equipment operating status and environmental parameters.

[0117] Data cleaning and preprocessing: Clean and preprocess the collected sensor data and drone inspection data to remove noise and outliers to ensure data quality. Fusion the cleaned sensor data and drone inspection data to form a comprehensive data set. Use deep learning algorithms to deeply analyze the data set and identify potential failure modes.

[0118] Based on the results of data fusion analysis, the specific location and type of the fault are determined, and corresponding maintenance suggestions are provided. Combined with information such as fault location, maintenance suggestions, and equipment status reports, a comprehensive inspection conclusion is generated. Based on the inspection conclusion, a specific fault repair plan is formulated, including emergency repair steps, preventive maintenance suggestions, etc.

[0119] Through the specific implementation methods of the above steps, the present invention can achieve efficient intelligent inspection, timely discover and handle equipment failures and potential problems, and ensure the stable operation of power plant equipment.

[0120] Those skilled in the art can clearly understand that the technology in the embodiments of the present invention can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution in the embodiments of the present invention can be essentially or partly contributed to the prior art in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present invention or certain parts of the embodiments.

[0121] The above-described embodiments of the present invention do not limit the protection scope of the present invention.

Claims

1. A power plant intelligent inspection system based on a large model, characterized in that: It includes the server side, the background control side, the inspection device side, the inspected device side and the user side. The server side establishes communication connections with the background control side, the inspection device side, the inspected device side and the user side. The server side is used to process and analyze the data transmitted by the background control side, the inspection device side, the inspected device side and the user side. The server side includes: The sensor fault detection unit obtains the data collected by the sensors on the intelligent inspection equipment and the power equipment, uses the random forest algorithm to train the data collected by the sensors, builds a sensor data model, predicts the sensor data through the sensor data model, obtains the model prediction data, compares the model prediction data with the real-time data, and identifies the data points that deviate from the normal range by calculating the deviation between the real-time data and the model prediction data. When multiple consecutive data points are marked as abnormal, the sensor fault warning information is issued. If the data points that deviate from the normal range are identified to be within the preset safety range, the data collected by the sensor are marked as usable detection data; The drone inspection optimization unit compares the available detection data with the preset safety values ​​of each device. If there is no abnormal data in the comparison result, the drone inspection will be carried out according to the preset path. If there is abnormal data in the comparison result, the location information corresponding to the abnormal data will be obtained and substituted into the drone inspection path generation model to generate the drone collaborative inspection path. The data fusion analysis unit receives and processes data from multiple sensors and drones for data fusion, uses deep learning algorithms to perform in-depth analysis on the fused data to identify potential faults, builds a decision support model, substitutes the identified potential faults into the decision support model, and generates information on fault location and maintenance suggestions for maintenance personnel; The data collection fault repair unit generates inspection conclusions through a pre-trained large model based on the fault location and repair suggestions provided by the maintenance personnel. The inspection conclusions include equipment status reports, abnormality lists and risk levels, and automatically generate inspection disposal plans; The re-inspection path generation unit generates the inspection path for the drone re-inspection based on the location information corresponding to the recorded abnormal data and the specific location of the fault repair. The data collected by the drone re-inspection is transmitted to the data fusion analysis unit.

2. A large model-based intelligent inspection system for power plants as claimed in claim 1, characterized in that: The sensor fault detection unit further comprises: Obtain the operation and management data of the power plant, including historical inspection records, equipment status, fault reports, and environmental monitoring data, clean the data to remove errors and outliers, and annotate the data to distinguish between normal, warning, and fault states of the equipment; Data features are extracted from the operation and maintenance and management data of the power plant, and the extracted data features are divided into training set, validation set and test set. The data features in the training set, validation set and test set are divided into time windows, and anomaly detection is performed to identify outliers in the data.

3. The large model-based intelligent inspection system for power plants according to claim 1, characterized in that: Data fusion analysis unit, including; Train the big model based on the data generated in the sensor fault detection unit and define the architecture of the big model, including the configuration of the attention mechanism, the number of encoder and decoder layers, and the number of heads; The preprocessed data set generated in the sensor fault detection unit is fed into the large model, and the self-attention mechanism is used to capture the long-range dependencies in the data; Through the optimizer and learning rate scheduling strategy, the training process of the large model is started to obtain the trained large model.

4. A large model-based intelligent inspection system for power plants as claimed in claim 3, characterized in that: UAV inspection optimization unit, including: Input the power plant's layout information, equipment distribution, historical fault data, and inspection record data into the trained large model for analysis; The trained large model combines machine learning and optimization algorithms to analyze the criticality, failure rate, interdependence and periodic inspection requirements of equipment, and generates one or more optimized inspection paths.

5. The large model-based intelligent inspection system for power plants according to claim 1, characterized in that: A sensor fault detection unit, comprising: Use the trained big model to analyze the data collected by the drone to automatically identify abnormalities and potential safety risks of the inspected equipment. The trained big model conducts in-depth analysis of the data and provides fault diagnosis and risk assessment for the inspected equipment. The image recognition algorithm is used to detect abnormal conditions on the inspected equipment, including cracks, corrosion, and foreign body intrusion. The thermal imaging analysis algorithm is used to identify potential faults such as overheating and insulation damage of the inspected equipment. Combined with the historical inspection data and real-time monitoring results, the trained large model is used to predict the failure trend of the inspected equipment.

6. The large model-based intelligent inspection system for power plants according to claim 1, characterized in that: The data acquisition fault repair unit also includes: The inspection and disposal plan is used to guide power maintenance personnel in carrying out maintenance and repair work. The inspection and disposal plan includes inspection problem description, location positioning, severity assessment and fault cause analysis.

7. The large model-based intelligent inspection system for power plants according to claim 1, characterized in that: The data acquisition fault repair unit also includes: Match the exception list and risk level in the historical database and knowledge base to obtain the cases and solutions corresponding to the exception list and risk level; Based on the current equipment status, available resources and the maintenance personnel's successful maintenance cases, the maintenance plan is matched to generate an inspection and disposal plan for the current fault. The inspection and disposal plan includes emergency repair steps, preventive maintenance suggestions and parts that need to be replaced.

8. The large model-based intelligent inspection system for power plants according to claim 7, characterized in that: The data acquisition fault repair unit also includes: Conduct preliminary screening of received exception lists and risk level information to remove duplicate, invalid or obviously erroneous data; Identify and determine abnormal data by comparing it with the normal parameter range of the abnormal list; According to the source of abnormal information, combined with the equipment layout and monitoring system data of the power plant, the specific location of the abnormal data is located.

9. The large model-based intelligent inspection system for power plants according to claim 8, characterized in that: The data acquisition fault repair unit also includes: Use abnormal information and risk level as search keywords to search in the historical database and find solutions and cases related to the current fault in the knowledge base; According to the similar fault cases in the historical records and the information in the knowledge base, the solutions to the abnormal information are matched to obtain the approximate solutions to the abnormal information.

10. The large model-based intelligent inspection system for power plants according to claim 8, characterized in that: The data acquisition fault repair unit also includes: Based on the approximate solution of abnormal information, the data format is converted to obtain the inspection and disposal plan. The inspection and disposal plan includes emergency repair steps, preventive maintenance suggestions, and information on parts that need to be replaced. The generated inspection and disposal plan is fed back to maintenance personnel.

Images