Electric power intelligent equipment operation and maintenance management method based on artificial intelligence

Abstract

The invention discloses a power intelligent equipment operation and maintenance management method based on artificial intelligence, and the method employs a staged training strategy, combines a LSTM and CNN mixed structure to carry out fault prediction, effectively captures an equipment operation state change trend, recognizes a potential fault mode, remarkably improves the fault prediction precision, and meanwhile, improves the fault prediction efficiency. A flexible self-healing strategy is designed according to the fault type and severity, the adaptability of a self-healing process and the capacity of dealing with complex faults are enhanced, in addition, a trained equipment health assessment, fault prediction and self-healing model is packaged into RESTfu API service in a micro-service mode and is deployed to an edge computing module through Docker containerization, and the service performance of the edge computing module is improved. Intellectualization and integration of the operation and maintenance management system are promoted, the edge calculation module inferes and monitors the equipment state in real time and generates early warning information, real-time monitoring and early warning are achieved, once abnormity is found, self-healing operation is triggered immediately, manual intervention is reduced, and the operation and maintenance efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of operation and maintenance management of electric power intelligent equipment, and in particular to an operation and maintenance management method of electric power intelligent equipment based on artificial intelligence. Background Art

[0002] With the continuous development of power systems and the improvement of their intelligence level, the application of smart power devices in power grids is becoming increasingly widespread. These devices not only undertake the important tasks of power transmission, distribution and conversion, but also realize real-time monitoring and intelligent control of the power grid operation status by integrating advanced sensors, communication technologies and control algorithms. However, with the increasing complexity of equipment and the diversification of operating environments, the operation and maintenance management of smart power devices faces unprecedented challenges.

[0003] Traditional operation and maintenance management of smart power equipment mainly relies on manual inspections and regular maintenance. This method is not only inefficient, but also makes it difficult to grasp the operating status and potential faults of the equipment in real time. Once the equipment fails, it often causes power outages, which have a serious impact on the operation of the power grid. Therefore, how to realize intelligent operation and maintenance management of smart power equipment and improve the reliability and operating efficiency of the equipment has become an urgent problem to be solved in the current power industry. Summary of the Invention

[0004] In view of this, the present invention proposes an artificial intelligence-based power intelligent equipment operation and maintenance management method, which can effectively solve the defects of the existing technology that relies on manual inspection and regular maintenance.

[0005] The technical solution of the present invention is achieved as follows:

[0006] An artificial intelligence-based power intelligent equipment operation and maintenance management method, specifically comprising:

[0007] Collect electrical parameter data and non-electrical parameter data of power intelligent equipment;

[0008] Preprocess the collected electrical parameter data and non-electrical parameter data, and upload the preprocessed data to the cloud platform;

[0009] Based on the uploaded data, the cloud platform adopts a phased training strategy to train the equipment health assessment model, fault prediction model, and fault self-healing model. The equipment health assessment model uses a multi-layer perceptron structure to calculate the equipment health index. The fault prediction model uses a hybrid structure of LSTM and CNN to predict the future operating status of the equipment and identify potential failure modes. The fault self-healing model is used to design corresponding self-healing strategies based on the fault type and severity.

[0010] The trained device health assessment model, fault prediction model, and fault self-healing model are encapsulated as RESTful API services in the form of microservices and deployed to the edge computing module of the power intelligent device through Docker containerization;

[0011] The edge computing module infers the real-time data, monitors the device status, and generates warning information according to the preset thresholds;

[0012] When receiving the warning information, it automatically triggers the fault self-healing process, selects the self-healing strategy with the highest matching degree from the self-healing strategy library according to the fault characteristics, and then performs security verification and feasibility verification before executing the self-healing strategy. After confirmation, the self-healing operation is executed. During the execution process, the device status is monitored in real time. Once an abnormality is found, the self-healing operation is immediately aborted and the user is notified. After the self-healing operation is completed, the self-healing effect is evaluated and the result is fed back to the cloud platform for optimizing the self-healing strategy. Among them, the self-healing strategies include parameter adjustment, load transfer, device restart, and standby device switching.

[0013] As a further optional solution of the power intelligent device operation and maintenance management method based on artificial intelligence, the data preprocessing of the collected electrical parameter data and non-electrical parameter data specifically includes:

[0014] Based on the 3σ principle, outlier detection is performed on the electrical parameter data and non-electrical parameter data, and the detected outliers are replaced with the average value of adjacent normal values to obtain the parameter data after outlier processing; [[ID=…]]

[0015] The sliding average method is used to perform noise reduction processing on the parameter data after outlier processing to obtain the parameter data after noise reduction processing;

[0016] Based on the parameter data after noise reduction processing, key statistical features that can reflect the device operation status are extracted to obtain feature data;

[0017] Huffman coding is used to compress the feature data to obtain the preprocessed data.

[0018] As a further optional solution of the power intelligent device operation and maintenance management method based on artificial intelligence, during the training process of the device health assessment model, fault prediction model, and fault self-healing model, the knowledge distillation technology is adopted to compress the knowledge of the complex model into a lightweight model.

[0019] As a further optional solution of the power intelligent device operation and maintenance management method based on artificial intelligence, the encapsulation of the trained device health assessment model, fault prediction model, and fault self-healing model as RESTful API services in the form of microservices specifically includes:

[0020] Each trained model is encapsulated as an independent microservice, and each microservice provides a set of RESTful API interfaces for receiving external requests and returning model inference results. Among them, for the device health assessment model, it is encapsulated as a service providing a health index query interface, which receives the device ID and related parameters and returns the device's health index and status assessment results. For the fault prediction model, it is encapsulated as a service providing a fault probability query interface, which receives the device ID and parameters of the prediction time window and returns the fault probabilities of the device at different future time points. For the fault self-healing model, it is encapsulated as a service providing a self-healing strategy query and execution interface, which receives the parameters of the fault type and device status and returns the recommended self-healing strategy and execution results.

[0021] As a further optional solution of the power intelligent device operation and maintenance management method based on artificial intelligence, the deployment to the edge computing module of the power intelligent device through Docker containerization specifically includes:

[0022] Build corresponding Docker images for each encapsulated microservice. The images include the running environment, dependency libraries, and model files required by the microservice;

[0023] Deploy the built Docker images to the edge computing module of the power intelligent device through Docker containerization technology;

[0024] Register the microservices deployed on the edge computing module to the service registry.

[0025] As a further optional solution of the power intelligent device operation and maintenance management method based on artificial intelligence, the method further includes fault attribution analysis based on a knowledge graph, specifically including:

[0026] Based on the preprocessed data, build a power intelligent device fault knowledge graph. The knowledge graph includes device entities, fault entities, attributes, relationships, and rules, and is used to represent the association relationships between devices, the propagation paths of faults, and the associations between faults and device states and environmental factors;

[0027] Use real-time monitoring data and historical fault patterns to detect the occurrence of faults through inference by the edge computing module, and locate the faulty device based on the association relationships in the knowledge graph;

[0028] Perform multi-source data fusion based on the knowledge graph, and integrate multi-source information such as device operation data, environmental data, historical fault records, and expert experience;

[0029] Adopt a change point detection algorithm to determine the starting time of the fault and extract the fault feature vector;

[0030] Combined with the knowledge graph and fault feature vectors after multi-source data fusion, conduct fault attribution analysis to determine the root cause of the fault.

[0031] A power intelligent device operation and maintenance management system based on artificial intelligence, comprising:

[0032] A data acquisition module for collecting electrical parameter data and non-electrical parameter data of power intelligent devices;

[0033] A data preprocessing module for preprocessing the collected electrical parameter data and non-electrical parameter data and uploading the preprocessed data to the cloud platform;

[0034] A cloud platform training module for training a device health assessment model, a fault prediction model, and a fault self-healing model based on the uploaded data using a phased training strategy. Among them, the device health assessment model adopts a multi-layer perceptron structure for calculating the device health index, the fault prediction model adopts a hybrid structure of LSTM and CNN for predicting the future operating state of the device and identifying potential fault modes, and the fault self-healing model is used to design corresponding self-healing strategies according to the fault type and severity;

[0035] A model encapsulation and deployment module for encapsulating the trained device health assessment model, fault prediction model, and fault self-healing model into RESTful API services in the form of microservices and deploying them to the edge computing module of power intelligent devices through Docker containerization technology;

[0036] An edge computing inference module for inferring real-time data, monitoring the device status, and generating warning information according to a preset threshold;

[0037] A fault self-healing execution module for automatically triggering the fault self-healing process when receiving a warning message, selecting the self-healing strategy with the highest matching degree from the self-healing strategy library according to the fault characteristics, then performing security verification and feasibility verification before executing the self-healing strategy, executing the self-healing operation after confirmation, monitoring the device status in real time during the execution process, immediately aborting the self-healing operation and notifying the user once an abnormality is found, evaluating the self-healing effect after the self-healing operation is completed and feedbacking the result to the cloud platform for optimizing the self-healing strategy. Among them, the self-healing strategies include parameter adjustment, load transfer, device restart, and standby device switching.

[0038] As a further optional solution of the power intelligent device operation and maintenance management system based on artificial intelligence, the system further includes:

[0039] The human-computer interaction module is used to provide a visual interface to display device status, warning information, fault handling suggestions, self-healing execution results, and system configuration parameters, and support remote configuration and monitoring functions, allowing users to view real-time data and historical records, receive warning notifications, execute remote commands, and modify system parameters through a remote interface.

[0040] A computing device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of any one of the above-mentioned power intelligent device operation and maintenance management methods based on artificial intelligence.

[0041] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, it implements the steps of any one of the above-mentioned power intelligent device operation and maintenance management methods based on artificial intelligence.

[0042] The beneficial effects of the present invention are as follows: By adopting a phased training strategy and combining a hybrid structure of LSTM and CNN for fault prediction, it can more accurately capture the changing trend of the device operation state, identify potential fault modes, and improve the accuracy of fault prediction. Corresponding self-healing strategies are designed according to the fault type and severity, making the self-healing process more flexible and adaptable, and capable of coping with complex and changeable fault situations. The trained device health assessment model, fault prediction model, and fault self-healing model are encapsulated as RESTful API services in the form of microservices and deployed to the edge computing module of power intelligent devices through Docker containerization, promoting the intelligence and integration of the operation and maintenance management system. The edge computing module infers real-time data, monitors device status, and generates warning information, realizing real-time monitoring and warning of device status. By real-time monitoring the device status, once an anomaly is detected, the self-healing operation is immediately triggered, reducing the need for manual inspection and intervention and improving the operation and maintenance efficiency. After the self-healing operation is completed, the self-healing effect is evaluated, and the result is fed back to the cloud platform for optimizing the self-healing strategy, forming a closed-loop management and continuously improving the operation and maintenance management level. Before executing the self-healing strategy, safety verification and feasibility verification are carried out to ensure that the self-healing operation will not have a negative impact on the power grid operation. By optimizing the self-healing strategy, the impact of device faults on the power grid operation is reduced, ensuring the safety and stability of the power grid. Description of the Drawings

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0044] Figure 1 Schematic flowchart of a method for operation and maintenance management of power intelligent devices based on artificial intelligence according to the present invention;

[0045] Figure 2 Schematic diagram of the composition of a system for operation and maintenance management of power intelligent devices based on artificial intelligence according to the present invention;

[0046] Figure 3 Schematic diagram of the composition of a computing device according to the present invention. Detailed implementation manners

[0047] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0048] Refer to Figures 1 to 3 , a method for operation and maintenance management of power intelligent devices based on artificial intelligence, specifically including:

[0049] Collect electrical parameter data and non-electrical parameter data of power intelligent devices;

[0050] Perform data preprocessing on the collected electrical parameter data and non-electrical parameter data, and upload the preprocessed data to the cloud platform;

[0051] Based on the uploaded data, the cloud platform adopts a phased training strategy to train a device health assessment model, a fault prediction model, and a fault self-healing model. Among them, the device health assessment model adopts a multi-layer perceptron structure and is used to calculate the device health index. The fault prediction model adopts a hybrid structure of LSTM and CNN and is used to predict the future operating state of the device and identify potential fault modes. The fault self-healing model is used to design corresponding self-healing strategies according to the fault type and severity;

[0052] Package the trained device health assessment model, fault prediction model, and fault self-healing model into RESTful API services in the form of microservices, and deploy them to the edge computing module of power intelligent devices through Docker containerization;

[0053] The edge computing module performs inference on real-time data, monitors the device status, and generates warning information according to a preset threshold;

[0054] When a warning message is received, the automatic fault self-healing process is triggered. The self-healing strategy with the highest matching degree is selected from the self-healing strategy library according to the fault characteristics. Subsequently, before executing the self-healing strategy, safety verification and feasibility verification are carried out. After confirmation, the self-healing operation is executed. During the execution process, the device status is monitored in real time. Once an anomaly is found, the self-healing operation is immediately aborted and the user is notified. After the self-healing operation is completed, the self-healing effect is evaluated and the result is fed back to the cloud platform for optimizing the self-healing strategy. Among them, the self-healing strategies include parameter adjustment, load transfer, device restart, and standby device switching.

[0055] In this embodiment, a phased training strategy is adopted, and a hybrid structure of LSTM and CNN is combined for fault prediction, which can more accurately capture the change trend of the device operation state, identify potential fault modes, and improve the accuracy of fault prediction. Corresponding self-healing strategies are designed according to the fault type and severity, making the self-healing process more flexible and adaptable, and capable of coping with complex and changeable fault situations. The trained device health assessment model, fault prediction model, and fault self-healing model are encapsulated as RESTful API services in the form of microservices and deployed to the edge computing module of the power intelligent device through Docker containerization, promoting the intelligence and integration of the operation and maintenance management system. The edge computing module infers the real-time data, monitors the device status, and generates warning messages, realizing the real-time monitoring and warning of the device status. By monitoring the device status in real time, once an anomaly is found, the self-healing operation is immediately triggered, reducing the need for manual inspection and intervention and improving the operation and maintenance efficiency. After the self-healing operation is completed, the self-healing effect is evaluated, and the result is fed back to the cloud platform for optimizing the self-healing strategy, forming a closed-loop management and continuously improving the operation and maintenance management level. Before executing the self-healing strategy, safety verification and feasibility verification are carried out to ensure that the self-healing operation will not have a negative impact on the power grid operation. By optimizing the self-healing strategy, the impact of device faults on the power grid operation is reduced, ensuring the safety and stability of the power grid.

[0056] It should be noted that electrical parameters such as voltage, current, active power, reactive power, frequency, phase, and harmonics of the power intelligent device are collected, as well as non-electrical parameters such as the device status, ambient temperature and humidity, vibration, and sound of the power intelligent device. In addition, the device health assessment model adopts a multi-layer perceptron structure, and calculates the device health index based on factors such as device operation parameters, manufacturer, model, service life, and working environment. The fault prediction model adopts a hybrid structure of LSTM and CNN, predicts the future operation state of the device, identifies potential fault modes, and outputs the fault probabilities for the next 24 hours, 72 hours, and 168 hours.

[0057] Preferably, the data preprocessing of the collected electrical parameter data and non-electrical parameter data specifically includes:

[0058] Perform outlier detection on electrical parameter data and non - electrical parameter data based on the 3σ principle, and replace the detected outliers with the average of adjacent normal values to obtain the parameter data after outlier processing;

[0059] Use the moving average method to perform noise reduction processing on the parameter data after outlier processing to obtain the parameter data after noise reduction processing;

[0060] Extract key statistical features that can reflect the operating state of the device from the parameter data after noise reduction processing to obtain feature data;

[0061] Use Huffman coding to compress the feature data to obtain the pre - processed data.

[0062] In this embodiment, performing outlier detection on electrical parameter data and non - electrical parameter data based on the 3σ principle can effectively identify outliers in the data. These outliers may be caused by sensor failures, data transmission errors, or other external interferences. Replacing the detected outliers with the average of adjacent normal values can reduce the impact of outliers on subsequent data analysis and improve the accuracy and reliability of the data. Using the moving average method to perform noise reduction processing on the parameter data after outlier processing can smooth the fluctuations in the data and reduce the interference of noise on data analysis. The moving average method calculates the average value of data within a certain window to replace the original data, thereby reducing the impact of random noise. Extracting key statistical features that can reflect the operating state of the device from the parameter data after noise reduction processing, such as mean, variance, peak value, etc. These feature data can concisely represent the operating state of the device, reduce the data dimension, and improve the efficiency of data analysis. Using Huffman coding to compress the feature data can reduce the data storage space under the premise of ensuring data integrity. Huffman coding is a lossless compression algorithm that encodes data according to the frequency of data occurrence, so that data with a high occurrence frequency is represented by a shorter code, thus achieving data compression.

[0063] It should be noted that for the outlier detection step based on the 3σ principle, that is, calculating the mean and standard deviation of each data point, marking the data points that exceed the mean ± 3 times the standard deviation as outliers, and processing the outliers according to preset rules, such as replacing them with the average of adjacent normal values. In addition, for the feature extraction step, extract key statistical features that can reflect the operating state of the device, such as mean, variance, peak value, trend, etc.; finally, use Huffman coding or wavelet compression algorithm to compress the data to reduce the storage space and network transmission burden.

[0064] Preferably, during the training process of the device health assessment model, fault prediction model, and fault self - healing model, knowledge distillation technology is adopted to compress the knowledge of complex models into lightweight models.

[0065] In this embodiment, the knowledge distillation technology can transfer the knowledge of a complex model (usually called the teacher model) to a lightweight model (usually called the student model). Since the lightweight model has fewer parameters and lower computational complexity, it can give prediction results faster during the inference process, improving the running efficiency of the model. In the operation and maintenance management of power intelligent devices, real-time performance is a key factor. The lightweight model can evaluate the device health status, predict faults, and generate self-healing strategies faster, thus improving the response speed and efficiency of operation and maintenance management; the lightweight model requires fewer computing resources and storage resources compared to the complex model, which means that fewer hardware resources can be used when deploying the model, reducing the overall cost of the system. Power intelligent devices are usually deployed in environments with limited resources, such as edge computing nodes. Using a lightweight model can reduce the dependence on hardware resources, enabling the model to run on more types of devices, improving flexibility and scalability; the knowledge distillation technology uses specific loss functions and optimization strategies to ensure that the lightweight model can maintain high prediction performance while inheriting the knowledge of the complex model. This means that the lightweight model can achieve an accuracy similar to that of the complex model in aspects such as device health assessment, fault prediction, and self-healing strategy generation.

[0066] It should be noted that on the cloud platform, based on the preprocessed data, a complex and high-performance teacher model is first trained. This model uses deep learning architectures such as multi-layer perceptron, long short-term memory network, convolutional neural network, or their hybrid structures. During the training process, the knowledge distillation technology is introduced to distill the knowledge of the teacher model into a lightweight student model. The knowledge distillation includes using the output probability distribution or intermediate layer features of the teacher model as soft labels to train the student model, so that the student model has fewer parameters and lower computational complexity while maintaining high accuracy. The student model is optimized and trained using the cross-entropy loss function or other suitable loss functions, combined with the knowledge distillation loss function, to ensure that the student model can accurately simulate the behavior of the teacher model. At the same time, multi-dimensional evaluation metrics such as accuracy, precision, recall, F1 score, and AUC value are used to evaluate the student model to ensure that its performance meets the actual application requirements.

[0067] Preferably, encapsulating the trained device health assessment model, fault prediction model, and fault self-healing model into RESTful API services in the form of microservices specifically includes:

[0068] Each trained model is encapsulated as an independent microservice, and each microservice provides a set of RESTful API interfaces for receiving external requests and returning model inference results. Among them, for the device health assessment model, it is encapsulated as a service providing a health index query interface, which receives the device ID and related parameters and returns the device's health index and status assessment results. For the fault prediction model, it is encapsulated as a service providing a fault probability query interface, which receives the device ID and parameters of the prediction time window and returns the fault probabilities of the device at different future time points. For the fault self-healing model, it is encapsulated as a service providing a self-healing strategy query and execution interface, which receives parameters such as the fault type and device status and returns the recommended self-healing strategy and execution results.

[0069] In this embodiment, each model is encapsulated as an independent microservice, achieving modularization and decoupling of functions. Each microservice is only responsible for specific functions, such as health assessment, fault prediction, or self-healing strategy generation. This makes the system clearer, easier to manage and maintain. In the operation and maintenance management of power intelligent devices, this modular design enables the system to be more easily extended and upgraded. For example, when a certain model needs to be updated, only the corresponding microservice needs to be updated without affecting the functions of other parts; each microservice provides a set of RESTful API interfaces for receiving external requests and returning model inference results. This standardized interface design makes it easier for different systems to be integrated and interact. Through the RESTful API interfaces, other systems (such as monitoring platforms, data analysis tools, etc.) can conveniently call the services of these models to achieve data sharing and collaborative work, which helps to build a more open and flexible operation and maintenance management system for power intelligent devices; the microservice architecture makes the system have higher flexibility and scalability, and microservices can be dynamically added or deleted according to actual needs to adapt to different operation and maintenance management scenarios. In the operation and maintenance management of power intelligent devices, different types of devices and fault situations may be encountered. Through the microservice architecture, the functions and services of the system can be flexibly adjusted to meet specific operation and maintenance needs; since each microservice runs independently and communicates through RESTful API interfaces, the system can respond to external requests faster and provide real-time model inference results. In the operation and maintenance management of power intelligent devices, real-time performance is a key factor. Through the microservice architecture, it can be ensured that key functions such as device health assessment, fault prediction, and self-healing strategy generation can provide results in real time, providing timely technical support for operation and maintenance management; for the device health assessment model, a health index query interface is provided, enabling operation and maintenance personnel to quickly obtain the health status and index of a specific device, facilitating timely maintenance measures. For the fault prediction model, a fault probability query interface is provided, allowing operation and maintenance personnel to obtain the fault probabilities at different future time points according to the device ID and prediction time window, which helps to plan maintenance plans and resource allocation in advance. For the fault self-healing model, self-healing strategy query and execution interfaces are provided, enabling operation and maintenance personnel to obtain recommended self-healing strategies according to the fault type and device status and directly execute or further adjust them, improving the efficiency and accuracy of fault handling.

[0070] Preferably, the deployment to the edge computing module of the power intelligent device through Docker containerization specifically includes:

[0071] Build corresponding Docker images for each encapsulated microservice. The images include the operating environment, dependent libraries, and model files required by the microservice;

[0072] Deploy the built Docker image to the edge computing module of the power intelligent device through Docker containerization technology;

[0073] Register the microservices deployed on the edge computing module to the service registry.

[0074] In this embodiment, corresponding Docker images are built for each encapsulated microservice. The images contain the running environment, dependency libraries, and model files required by the microservices, which ensures the consistent operation of the microservices in different environments and avoids compatibility issues caused by environmental differences. At the same time, the portability of Docker images enables microservices to be easily deployed to different power intelligent devices or edge computing nodes. In the operation and maintenance management of power intelligent devices, the diversity and distribution of devices are a challenge. Through Docker containerization technology, it can be ensured that microservices can run stably in any environment that supports Docker, improving portability and deployment efficiency; Docker containers provide a lightweight resource isolation mechanism, enabling each microservice to run in an independent container without interference, which improves the security of the system and prevents problems in one microservice from affecting other services. In the edge computing environment, resources are limited and security requirements are high. Through Docker containerized deployment, it can be ensured that each microservice has sufficient resources to run while reducing security risks; Docker containerization technology makes the deployment of microservices fast and simple. Just deploy the built Docker image to the target device to start the microservice. In addition, when the system needs to be scaled, just add more container instances without complex configuration and modification of the entire system; Registering the microservices deployed on the edge computing module to the service registry enables other services or clients to easily discover and call these microservices, which improves the flexibility and maintainability of the system and makes it easier to add, delete, and update services. In the operation and maintenance management system of power intelligent devices, there may be multiple microservices and clients that need to communicate with each other. Through the service registry, the registration and discovery of these services can be dynamically managed to ensure that all parts of the system can work together.

[0075] Preferably, the method further includes performing fault attribution analysis based on a knowledge graph, specifically including:

[0076] Based on the preprocessed data, construct a power intelligent device fault knowledge graph. The knowledge graph includes device entities, fault entities, attributes, relationships, and rules, and is used to represent the association relationships between devices, the propagation paths of faults, and the associations between faults and device states and environmental factors;

[0077] Using real-time monitoring data and historical fault patterns, the edge computing module infers and detects the occurrence of faults, and locates the faulty devices based on the association relationships in the knowledge graph;

[0078] Based on the knowledge graph, multi-source data fusion is carried out to integrate multi-source information such as device operation data, environmental data, historical fault records, and expert experience;

[0079] The change point detection algorithm is used to determine the starting time of the fault and extract the fault feature vector;

[0080] Combining the knowledge graph after multi-source data fusion and the fault feature vector, fault attribution analysis is carried out to determine the root cause of the fault.

[0081] In this embodiment, by constructing a fault knowledge graph of power intelligent devices, elements such as device entities, fault entities, attributes, relationships, and rules are integrated together to form a comprehensive representation of device fault knowledge. This representation method can clearly show the association relationships between devices, the propagation paths of faults, and the complex associations between faults and device states and environmental factors. The comprehensive fault knowledge representation helps operation and maintenance personnel better understand the essence and occurrence mechanism of faults, providing a solid foundation for fault attribution analysis; using real-time monitoring data and historical fault patterns, the edge computing module infers and detects the occurrence of faults, and locates the faulty device based on the association relationships in the knowledge graph. This location method combines real-time data and historical experience, improving the accuracy and efficiency of fault location. Accurate fault location is the key to quickly responding to and repairing faults. Through the association relationships in the knowledge graph, the faulty device can be quickly located, reducing the time for fault troubleshooting; based on the knowledge graph for multi-source data fusion, integrating multi-source information such as device operation data, environmental data, historical fault records, and expert experience. This fusion method can make full use of various data resources, improving the comprehensiveness and accuracy of fault attribution analysis. Multi-source data fusion helps operation and maintenance personnel understand the fault situation from multiple perspectives, avoiding the one-sidedness that may be brought by a single data source. Through the integration function of the knowledge graph, a comprehensive understanding of the fault can be formed; the change point detection algorithm is used to determine the starting time of the fault and extract the fault feature vector. This algorithm can accurately identify the change points in the data, that is, the starting time of the fault occurrence, providing accurate time information for fault attribution analysis. The accurate starting time of the fault is crucial for fault attribution analysis. Through the change point detection algorithm, the time point of the fault occurrence can be accurately determined; combining the knowledge graph after multi-source data fusion and the fault feature vector, fault attribution analysis is carried out to determine the root cause of the fault. This analysis method can deeply explore the root cause behind the fault, providing a basis for formulating effective repair measures and prevention strategies. In-depth fault attribution analysis helps operation and maintenance personnel understand the essence and occurrence mechanism of faults, so as to formulate targeted repair measures and prevention strategies. Through the combination of the knowledge graph and the fault feature vector, accurate fault attribution can be achieved.

[0082] Preferably, an operation and maintenance management system for power intelligent devices based on artificial intelligence includes:

[0083] A data acquisition module for acquiring electrical parameter data and non-electrical parameter data of power intelligent devices;

[0084] A data preprocessing module for preprocessing the acquired electrical parameter data and non-electrical parameter data and uploading the preprocessed data to the cloud platform;

[0085] A cloud platform training module, which is used to train a device health assessment model, a fault prediction model, and a fault self-healing model based on the uploaded data by adopting a phased training strategy. Among them, the device health assessment model adopts a multi-layer perceptron structure and is used to calculate the device health index. The fault prediction model adopts a hybrid structure of LSTM and CNN and is used to predict the future operating state of the device and identify potential fault modes. The fault self-healing model is used to design corresponding self-healing strategies according to the fault type and severity;

[0086] A model encapsulation and deployment module, which is used to encapsulate the trained device health assessment model, fault prediction model, and fault self-healing model into RESTful API services in the form of microservices, and deploy them to the edge computing module of the power intelligent device through Docker containerization technology;

[0087] An edge computing inference module, which is used to perform inferences on real-time data, monitor the device status, and generate warning information according to preset thresholds;

[0088] A fault self-healing execution module, which is used to automatically trigger the fault self-healing process when receiving warning information, select the self-healing strategy with the highest matching degree from the self-healing strategy library according to the fault characteristics, and then perform security verification and feasibility verification before executing the self-healing strategy. After confirmation, execute the self-healing operation. During the execution process, monitor the device status in real time. Once an abnormality is found, immediately abort the self-healing operation and notify the user. After the self-healing operation is completed, evaluate the self-healing effect and feedback the result to the cloud platform for optimizing the self-healing strategy. Among them, the self-healing strategies include parameter adjustment, load transfer, device restart, and standby device switching.

[0089] Preferably, the system further includes:

[0090] A human-computer interaction module, which is used to provide a visual interface, display the device status, warning information, fault handling suggestions, self-healing execution results, and system configuration parameters, and support remote configuration and monitoring functions, allowing users to view real-time data and historical records, receive warning notifications, execute remote commands, and modify system parameters through the remote interface.

[0091] A computing device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of any one of the above-mentioned artificial intelligence-based power intelligent device operation and maintenance management methods are implemented.

[0092] A computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the above-mentioned artificial intelligence-based power intelligent device operation and maintenance management methods are implemented.

[0093] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An operation and maintenance management method for power intelligent devices based on artificial intelligence, characterized in that, Specifically, it includes: Collecting electrical parameter data and non-electrical parameter data of power intelligent devices; Performing data preprocessing on the collected electrical parameter data and non-electrical parameter data, and uploading the preprocessed data to the cloud platform; Based on the uploaded data, the cloud platform adopts a phased training strategy to train a device health assessment model, a fault prediction model, and a fault self-healing model. Among them, the device health assessment model adopts a multi-layer perceptron structure to calculate the device health index. The fault prediction model adopts a hybrid structure of LSTM and CNN to predict the future operating state of the device and identify potential fault modes. The fault self-healing model is used to design corresponding self-healing strategies according to the fault type and severity; Encapsulating the trained device health assessment model, fault prediction model, and fault self-healing model into RESTful API services in the form of microservices, and deploying them to the edge computing module of power intelligent devices through Docker containerization; The edge computing module performs inference on real-time data, monitors the device status, and generates warning information according to preset thresholds; When receiving the warning information, automatically trigger the fault self-healing process, select the self-healing strategy with the highest matching degree from the self-healing strategy library according to the fault characteristics, and then perform security verification and feasibility verification before executing the self-healing strategy. After confirmation, execute the self-healing operation. During the execution process, monitor the device status in real time. Once an abnormality is found, immediately abort the self-healing operation and notify the user. After the self-healing operation is completed, evaluate the self-healing effect and feedback the result to the cloud platform for optimizing the self-healing strategy. Among them, the self-healing strategies include parameter adjustment, load transfer, device restart, and standby device switching.

2. The method for operation and maintenance management of a power intelligent device based on artificial intelligence according to claim 1, wherein, The data preprocessing of the collected electrical parameter data and non-electrical parameter data specifically includes: Performing outlier detection on the electrical parameter data and non-electrical parameter data based on the 3σ principle, and replacing the detected outliers with the average value of adjacent normal values to obtain parameter data after outlier processing; Performing noise reduction processing on the parameter data after outlier processing using a moving average method to obtain parameter data after noise reduction processing; Extracting key statistical features that can reflect the device operating state from the parameter data after noise reduction processing to obtain feature data; Compressing the feature data using Huffman coding to obtain preprocessed data.

3. The method for operation and maintenance management of power intelligent devices based on artificial intelligence according to claim 2, characterized in that, During the training process of training the device health assessment model, fault prediction model, and fault self-healing model, knowledge distillation technology is adopted to compress the knowledge of complex models into lightweight models.

4. The method for operation and maintenance management of power intelligent devices based on artificial intelligence according to claim 3, characterized in that, The encapsulation of the trained device health assessment model, fault prediction model, and fault self-healing model into RESTful API services in the form of microservices specifically includes: Each trained model is encapsulated as an independent microservice, and each microservice provides a set of RESTful API interfaces for receiving external requests and returning model inference results. Among them, for the device health assessment model, it is encapsulated as a service providing a health index query interface, which receives the device ID and related parameters and returns the device's health index and status assessment results. For the fault prediction model, it is encapsulated as a service providing a fault probability query interface, which receives the device ID and parameters of the prediction time window and returns the fault probabilities of the device at different future time points. For the fault self-healing model, it is encapsulated as a service providing self-healing strategy query and execution interfaces, which receives the parameters of the fault type and device status and returns the recommended self-healing strategy and execution results.

5. A method for operation and maintenance management of power intelligent devices based on artificial intelligence according to claim 4, characterized in that, The deployment to the edge computing module of the power intelligent device through Docker containerization specifically includes: Construct corresponding Docker images for each encapsulated microservice, and the images include the operating environment, dependency libraries, and model files required by the microservice; Deploy the constructed Docker images to the edge computing module of the power intelligent device through Docker containerization technology; Register the microservices deployed on the edge computing module to the service registry center.

6. A method for operation and maintenance management of power intelligent devices based on artificial intelligence according to claim 5, characterized in that, The method further includes fault attribution analysis based on the knowledge graph, specifically including: Based on the preprocessed data, construct a fault knowledge graph of power intelligent devices. The knowledge graph includes device entities, fault entities, attributes, relationships, and rules, and is used to represent the association relationships between devices, the propagation paths of faults, and the associations between faults and device status and environmental factors; Use real-time monitoring data and historical fault patterns to detect the occurrence of faults through inference by the edge computing module, and locate the faulty device based on the association relationships in the knowledge graph; Perform multi-source data fusion based on the knowledge graph, and integrate multi-source information such as device operation data, environmental data, historical fault records, and expert experience; Adopt a change point detection algorithm to determine the starting time of the fault and extract the fault feature vector; Combine the knowledge graph after multi-source data fusion and the fault feature vector to conduct fault attribution analysis and determine the root cause of the fault.

7. An operation and maintenance management system for power intelligent devices based on artificial intelligence, characterized in that, It includes: A data acquisition module for collecting electrical parameter data and non-electrical parameter data of power intelligent devices; A data preprocessing module for preprocessing the collected electrical parameter data and non-electrical parameter data and uploading the preprocessed data to the cloud platform; A cloud platform training module for training a device health assessment model, a fault prediction model, and a fault self-healing model based on the uploaded data using a phased training strategy. Among them, the device health assessment model adopts a multi-layer perceptron structure for calculating the device health index, the fault prediction model adopts a hybrid structure of LSTM and CNN for predicting the future operating state of the device and identifying potential fault patterns, and the fault self-healing model is used to design corresponding self-healing strategies according to the fault type and severity. The model encapsulation and deployment module is used to encapsulate the trained device health assessment model, fault prediction model, and fault self-healing model into RESTful API services in the form of microservices, and deploy them to the edge computing module of power intelligent devices through Docker containerization technology; The edge computing inference module is used to infer real-time data, monitor the device status, and generate warning information according to preset thresholds; The fault self-healing execution module is used to automatically trigger the fault self-healing process when receiving warning information, select the self-healing strategy with the highest matching degree from the self-healing strategy library according to the fault characteristics, and then perform security verification and feasibility verification before executing the self-healing strategy. After confirmation, execute the self-healing operation. During the execution process, monitor the device status in real time. Once an abnormality is found, immediately abort the self-healing operation and notify the user. After the self-healing operation is completed, evaluate the self-healing effect and feedback the result to the cloud platform for optimizing the self-healing strategy. Among them, the self-healing strategies include parameter adjustment, load transfer, device restart, and standby device switching.

8. An operation and maintenance management system for power intelligent devices based on artificial intelligence according to claim 7, characterized in that, The system further includes: The human-computer interaction module is used to provide a visual interface to display device status, warning information, fault handling suggestions, self-healing execution results, and system configuration parameters, and support remote configuration and monitoring functions, allowing users to view real-time data and historical records, receive warning notifications, execute remote commands, and modify system parameters through the remote interface.

9. A computing device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the artificial intelligence-based operation and maintenance management method for power intelligent devices according to any one of claims 1-6.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the storage medium. When the computer program is executed by the processor, it implements the steps of the artificial intelligence-based operation and maintenance management method for power intelligent devices according to any one of claims 1-6.

Images