Electric power AI intelligent auxiliary control and inspection system

Abstract

The invention discloses an electric power AI intelligent auxiliary control and inspection system, and relates to the technical field of electric power, and the system comprises a data collection module which is used for collecting the operation state data and appearance image data of electric power equipment, and carrying out the preprocessing; the image processing and analysis module is used for generating an appearance fault prediction result; the health assessment and fault prediction module is used for generating an equipment health state and an operation fault prediction result; the inspection and control module is used for scheduling and executing inspection tasks, the inspection tasks comprise ground equipment inspection and power line inspection, the ground equipment inspection is executed through a robot, the power line inspection is executed through an unmanned aerial vehicle, and the inspection tasks are scheduled according to priorities; and the data security and transparency module is used for encrypting and storing all data in the system through the block chain. According to the invention, comprehensive monitoring of the operation state of the power equipment, accurate prediction of faults and efficient inspection can be realized, and the security and transparency of data are ensured at the same time.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power, and more particularly, to a power AI intelligent auxiliary control and inspection system. Background Art

[0002] With the continuous expansion of the scale of the power system, the number of power equipment and the maintenance requirements are also increasing day by day. The traditional inspection and maintenance methods of power equipment mainly rely on manual operation, which has problems such as low efficiency, high cost, and great potential safety hazards. Especially in complex environments and bad weather conditions, the difficulty and risk of manual inspection are further increased. In addition, the failures of power equipment often have suddenness and concealment, and the traditional regular inspection methods are difficult to achieve real-time monitoring and early warning, resulting in relatively large economic losses and social impacts after the failures occur.

[0003] In recent years, the rapid development of artificial intelligence (AI) technology has provided new solutions for the intelligent auxiliary control and inspection of power equipment. By introducing AI technology, it is possible to achieve real-time monitoring of the operating status of power equipment, fault prediction, and intelligent inspection, thereby greatly improving the safety and reliability of the power system. However, there are still some deficiencies in the existing power equipment inspection systems:

[0004] 1) Existing systems usually only rely on single sensor data (such as current, voltage), which is difficult to comprehensively reflect the operating status and appearance of the equipment, resulting in insufficient accuracy of fault prediction.

[0005] 2) Existing systems mostly use traditional statistical models or simple machine learning algorithms for fault prediction, which are difficult to handle complex non-linear relationships and time series data, and the prediction accuracy is limited.

[0006] 3) Existing inspection systems mostly rely on manual operation or simple automated equipment, which are difficult to adapt to complex environments and changing task requirements, and the inspection efficiency is low.

[0007] 4) The data storage and transmission of existing systems mostly use traditional methods, which have the risk of data leakage and tampering, and it is difficult to meet the high requirements of the power system for data security. Summary of the Invention

[0008] The present invention aims to provide a power AI intelligent auxiliary control and inspection system, which can achieve comprehensive monitoring of the operating status of power equipment, accurate prediction of faults, and efficient inspection through multi-source data acquisition, deep learning models, intelligent inspection equipment, and blockchain technology, while ensuring the security and transparency of data.

[0009] The technical solution adopted by the present invention is as follows:

[0010] The present invention provides a power AI intelligent auxiliary control and inspection system, comprising:

[0011] A data acquisition module, which is used to collect the operation status data and appearance image data of power equipment through current sensors, voltage sensors, temperature sensors, humidity sensors, infrared thermal imaging cameras and visual image sensors, and perform preprocessing;

[0012] An image processing and analysis module, which is used to perform image processing on the basis of a convolutional neural network model according to the appearance image data, identify whether there are appearance faults or potential hidden dangers in the device appearance, and generate an appearance fault prediction result;

[0013] A health assessment and fault prediction module, which is used to perform real-time assessment of the device health status based on a long short-term memory network model according to the operation status data, obtain the device health status, and predict possible faults and their time windows according to the device historical operation status and the device current operation status, and generate an operation fault prediction result;

[0014] An inspection and control module, which is used to comprehensively judge the maintenance requirements of the device according to the device health status, operation fault prediction result, appearance fault prediction result and environmental data, schedule and execute inspection tasks. The inspection tasks include ground equipment inspection and power line inspection. The ground equipment inspection is executed by a robot, and the power line inspection is executed by a drone. The inspection tasks are scheduled according to the priority. After the inspection tasks are executed, the inspection results are fed back in real time, and corresponding control operations are performed on the device according to the feedback results;

[0015] A data security and transparency module, which is used to encrypt and store all data in the system through a blockchain, and can generate a data query report for authorized users to query. The data query report includes device operation status, historical fault records and inspection task execution records.

[0016] Further, the long short-term memory network model uses multiple hidden layers to learn and model time series data. When inputting data, the operation status data of the device is segmented by a sliding window method, the model is gradually trained, and based on the training results, the potential fault time window of the device is predicted.

[0017] Further, the convolutional neural network model uses the appearance image data to extract the temperature distribution characteristics of each part of the device surface through multiple convolutional kernels, and judge whether there are appearance faults or potential hidden dangers; the convolutional neural network model includes a convolutional layer, a pooling layer, and a fully connected layer. The specific judgment process includes:

[0018] In the first stage, feature extraction is performed on the input appearance image data through the convolutional layer to obtain the temperature change area in the device image;

[0019] In the second stage, downsampling is performed on the output of the convolutional layer through the pooling layer to extract high-order features;

[0020] In the third stage, high-order features are fused through the fully connected layer, and the classification result is output to determine whether there are appearance faults or potential hidden dangers in the device.

[0021] Furthermore, the inspection and control module manages tasks through the cloud scheduling system, specifically including:

[0022] According to the device health status and the operation fault prediction result, it is judged whether the device needs to be routinely inspected. If the device health degree is low or the fault prediction result indicates that a fault is about to occur, robots or drones are preferentially scheduled to inspect the device;

[0023] According to the appearance fault prediction result, it is judged whether there are external visible faults or potential hidden dangers in the device. If the appearance fault prediction result shows that the device has obvious appearance faults or hidden dangers, robots or drones are preferentially scheduled to check or repair the device;

[0024] Based on the device health status, the operation fault prediction result and the appearance fault prediction result, the priorities of each inspection task are dynamically adjusted;

[0025] During the task assignment process, according to the inspection task priorities and geographical distribution of the devices, the task assignment strategy is optimized through reinforcement learning to ensure that high-priority devices are preferentially inspected.

[0026] Furthermore, after the inspection task is executed, the drone or robot records the device status detected through image acquisition and sensor data, and uploads it to the cloud platform in real time. Then, it is processed by the intelligent analysis module to generate an inspection report, and based on the device historical fault data and predictive analysis, an automatic subsequent inspection plan is generated, giving a recommended solution for the next round of inspection; the inspection and control module can also trigger automatic repair or adjust the operation status of the power equipment in case of an emergency fault.

[0027] Furthermore, each data block in the blockchain network is verified through digital signatures, and the encrypted storage mechanism of the blockchain uses the SHA256 hash algorithm to perform hash processing on each data block.

[0028] Furthermore, the blockchain is based on the public-private key encryption mechanism, and uses the asymmetric encryption method to ensure that the device data will not be accessed or tampered with by unauthorized third parties during the transmission and storage processes; the public key is used to verify the authenticity of the data, and the private key is used for the encrypted storage of the data. All blockchain data is irreversibly encrypted and stored in the blockchain, and only authorized users can access and view it.

[0029] Furthermore, the data security and transparency module performs data access control through a permission management system, enabling users at different levels to access corresponding data and functions according to their authorized permissions. The permission management system includes a user identity authentication module and a permission allocation module. The user identity authentication module verifies users through multi-factor authentication methods, and the permission allocation module assigns different data access permissions according to the roles and responsibilities of different users.

[0030] Furthermore, the robot and the drone perform inspection tasks using autonomous navigation and obstacle avoidance methods. The autonomous navigation is based on the fusion of lidar and vision sensors for positioning and path planning. During the inspection process, the robot and the drone analyze the equipment status by combining the real-time obtained equipment operation data and environmental information with a deep learning model to determine whether emergency maintenance is required or the equipment operation status needs to be adjusted.

[0031] Furthermore, the inspection and control module comprehensively analyzes the equipment health status, operation fault prediction results, appearance fault prediction results, and environmental data through a multi-dimensional data fusion method. The multi-dimensional data fusion method includes Kalman filtering and particle filtering algorithms. Kalman filtering is used to smooth the sequential data of the equipment, and particle filtering is applied to non-linear systems to simulate multiple particle trajectories for state prediction and optimize the equipment health assessment results.

[0032] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0033] The operation status data and appearance image data of power equipment are comprehensively collected through current sensors, voltage sensors, temperature sensors, humidity sensors, infrared thermal imaging cameras, and visual image sensors, and preprocessed to provide a high-quality data basis for subsequent analysis and prediction.

[0034] Based on the convolutional neural network (CNN) model, the equipment appearance images are processed and analyzed, which can efficiently identify equipment appearance faults or potential hidden dangers, generate accurate appearance fault prediction results, and provide an important basis for equipment maintenance.

[0035] Based on the long short-term memory network (LSTM) model, the operation status data of the equipment are analyzed in real time, which can accurately evaluate the equipment health status, predict possible faults and their time windows, generate reliable operation fault prediction results, and help operation and maintenance personnel take preventive measures in advance to avoid faults.

[0036] By using the robot and the drone to perform ground equipment inspection and power line inspection respectively, it can adapt to complex environments and changing task requirements, greatly improve the inspection efficiency. At the same time, the inspection tasks are scheduled according to priorities to ensure that key equipment is maintained in a timely manner.

[0037] All data in the system are encrypted and stored through blockchain technology to ensure data security and immutability. At the same time, the generated data query reports are available for authorized users to query, improving data transparency and traceability.

[0038] Through intelligent fault prediction and inspection, the frequency and risk of manual inspection are reduced, the operation and maintenance costs are lowered, and at the same time, the safety and reliability of the power system are improved.

[0039] To make the above objects, features, and advantages of the present invention more obvious and understandable, specific embodiments of the present invention are hereinafter given, in conjunction with the accompanying drawings, and the detailed description is as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0041] Figure 1 is a schematic block diagram of a power AI intelligent auxiliary control and inspection system;

[0042] Figure 2 is a flowchart for generating appearance fault prediction results;

[0043] Figure 3 is a flowchart for generating operation fault prediction results. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them.

[0045] Please refer to Figure 1 , an embodiment of the present invention provides a power AI intelligent auxiliary control and inspection system, including a data acquisition module, an image processing and analysis module, a health assessment and fault prediction module, an inspection and control module, and a data security and transparency module.

[0046] 1. Data Acquisition Module

[0047] The data acquisition module is used to collect the operation state data and appearance image data of power equipment through current sensors, voltage sensors, temperature sensors, humidity sensors, infrared thermal imaging cameras, and visual image sensors.

[0048] The current sensor is installed in the current loop of the power equipment and is arranged at the current input and output ends of key equipment such as cable joints, circuit breakers, and transformers. A non-invasive clamp-type current sensor is adopted to avoid interfering with the normal operation of the power equipment. The number of current sensors is determined according to the number of current loops of the equipment, and one current sensor is arranged for each loop.

[0049] The voltage sensor is installed at the voltage measurement points of the power equipment and is arranged at the voltage input and output ends of equipment such as transformers, switch cabinets, and distribution boxes. A high-precision voltage transformer is adopted and directly connected to the voltage measurement points. The number of voltage sensors is determined according to the number of voltage measurement points of the equipment, and one voltage sensor is arranged for each measurement point.

[0050] The temperature sensor is installed at the key heat-generating parts of the power equipment, such as transformer windings, cable joints, and switch contacts. A contact-type temperature sensor (such as a thermocouple or thermal resistor) or a non-contact infrared temperature sensor is adopted. The number of temperature sensors is determined according to the hot spot distribution of the equipment, and one temperature sensor is arranged for each hot spot.

[0051] The humidity sensor is installed in the surrounding environment of the power equipment and is arranged in enclosed spaces such as switch cabinets, distribution rooms, and transformer rooms. A wall-mounted or embedded humidity sensor is adopted to ensure that the sensor can accurately reflect the environmental humidity. The number of humidity sensors is determined according to the area and humidity distribution of the environment where the equipment is located, and one humidity sensor is arranged for each enclosed space.

[0052] The infrared thermal imaging camera is installed near the key parts of the power equipment and is arranged around equipment such as transformers, switch cabinets, and cable joints. A fixed or movable infrared thermal imaging camera is adopted to ensure that the camera can cover the key heat-generating areas of the equipment. The number of infrared thermal imaging cameras is determined according to the number and distribution of the equipment, and one infrared thermal imaging camera is arranged for each key equipment.

[0053] The visual image sensor is installed outside the power equipment and is arranged on the outside of equipment such as transformers, switch cabinets, and cable joints. A high-resolution camera is adopted to ensure that the external image of the equipment can be clearly captured. The number of visual image sensors is determined according to the number and distribution of the equipment, and one visual image sensor is arranged for each key equipment.

[0054] For the current sensor and the voltage sensor, their acquisition methods are as follows: Real-time acquisition of current and voltage data, and the sampling frequency is dynamically adjusted according to the operating state of the equipment (for example, low-frequency sampling is adopted during normal operation, and high-frequency sampling is adopted during abnormal operation).

[0055] For temperature sensors and humidity sensors, their acquisition method is as follows: regularly acquire temperature and humidity data, and the sampling interval is dynamically adjusted according to environmental changes (for example, short interval sampling is used when the environment changes rapidly, and long interval sampling is used when the environment is stable).

[0056] For infrared thermal imaging cameras and visual image sensors, their acquisition method is as follows: acquire image data either regularly or in a triggered manner, and trigger the acquisition when the device operating status is abnormal or environmental conditions change.

[0057] In this embodiment, for sensors fixedly installed (such as current sensors, voltage sensors, temperature sensors, humidity sensors), industrial communication protocols such as RS485, Modbus, and CAN are used to transmit data to the local data acquisition terminal in a wired manner. For sensors installed in a mobile manner (such as infrared thermal imaging cameras, visual image sensors), wireless communication protocols such as Wi-Fi, ZigBee, and LoRa are used to transmit data to the local data acquisition terminal wirelessly. The local data acquisition terminal transmits the data to the cloud platform through Ethernet or 4G / 5G networks to ensure the real-time and reliability of the data.

[0058] In this embodiment, the acquired operating status data and appearance image data need to be preprocessed, including data cleaning, data normalization, and data compression. Data cleaning is to remove noise and outliers in the acquired data to ensure the accuracy of the data. Data normalization is to uniformly convert the data of different sensors into a standard format for subsequent processing and analysis. Data compression is to compress image data and video data to reduce the overhead of data transmission and storage.

[0059] When storing data, short-term data is stored in the local data acquisition terminal to ensure that the data can still be saved in case of network interruption, and the data is transmitted to the cloud platform for long-term storage for subsequent query and analysis.

[0060] Through the above layout method and data acquisition and transmission method, the data acquisition module can efficiently and accurately acquire the operating status data and appearance image data of power equipment, providing reliable data support for subsequent image processing, health assessment, fault prediction, and inspection control.

[0061] 2. Image Processing and Analysis Module

[0062] The image processing and analysis module is used to perform image processing based on the appearance image data using a convolutional neural network (CNN) model for image analysis, identify whether there are appearance faults or potential hidden dangers in the device appearance, and generate appearance fault prediction results.

[0063] In this embodiment, the convolutional neural network model uses appearance image data to extract the temperature distribution characteristics of each part of the device surface through multiple convolutional kernels, and determines whether there are appearance faults or potential hidden dangers; the convolutional neural network model includes a convolutional layer, a pooling layer, and a fully connected layer, as Figure 2 shown. The specific judgment process is as follows:

[0064] 1) Feature extraction is performed on the input appearance image data through the convolutional layer to obtain the temperature change area in the device image.

[0065] The convolutional layer receives the preprocessed appearance image data as input. The convolutional layer uses multiple convolutional kernels to perform sliding convolutional operations on the input image. Each convolutional kernel extracts different local features in the image. In the prediction of power equipment appearance faults, these features correspond to the temperature change areas of each part of the device surface. Through convolutional operations, each convolutional kernel generates a corresponding feature map. The feature map reflects the feature response intensity at different positions in the input image. The convolutional layer can extract rich feature information through multiple convolutional kernels, thus obtaining a set of feature maps.

[0066] 2) The feature maps output by the convolutional layer are downsampled through the pooling layer to extract high-order features.

[0067] Through the downsampling operation, the size of the feature map can be reduced, the computational amount can be reduced, and at the same time, important feature information can be retained. The pooling layer can not only reduce the dimension of the feature map, but also extract high-order features. During the pooling process, some local detail information is ignored, while more representative high-order features are retained. These high-order features can better reflect the overall appearance characteristics of power equipment and contribute to subsequent fault judgment.

[0068] 3) The high-order features are fused through the fully connected layer, and the classification result is output.

[0069] The fully connected layer receives the high-order features output by the pooling layer and expands them into a one-dimensional vector. Each neuron in the fully connected layer is connected to all neurons in the previous layer. Through the calculation of weights and biases, the high-order features are fused. Different position and different type feature information can be integrated, thus obtaining a more representative feature vector. The last layer of the fully connected layer uses the softmax activation function to convert the feature vector into classification probabilities. The classification result indicates whether there are appearance faults or potential hidden dangers in the power equipment.

[0070] 4) Determine whether there are appearance faults or potential hidden dangers in the device, and generate the appearance fault prediction result.

[0071] Specifically, if the classification result shows that the device has a fault or potential hazard, the type and severity of the fault are further analyzed. The judgment result is used to generate an appearance fault prediction result, which includes information such as the type, location, and severity of the fault, providing a basis for subsequent inspections and maintenance.

[0072] The convolutional neural network model is trained using the following loss function:

[0073]

[0074] where y i is the actual fault category, f(x i ; θ) is the predicted value of the model, N is the number of samples, θ is the model parameter, and the loss function is used to minimize the error between the prediction result and the true label.

[0075] 3. Health Assessment and Fault Prediction Module

[0076] As shown in combination with Figure 3 , the health assessment and fault prediction module is used to perform real-time assessment of the health status of the device based on the operation status data using a long short-term memory network (LSTM) model for time series analysis, obtain the device health status, and predict possible faults and their time windows based on the device's historical operation status and current operation status, generating an operation fault prediction result.

[0077] In this embodiment, based on the LSTM model, the operation status of the device (such as time series data such as current and voltage fluctuations) is analyzed to perform a health status assessment and output the health score of the device. The health score is calculated using the following formula:

[0078]

[0079] where H(t) is the health assessment value of the device at time point t, d i (t) is the diagnostic data of device i at time t, w i is the weight coefficient, N is the number of devices, and the health assessment value is weighted and averaged based on the health data of each device to comprehensively evaluate the overall status of the device.

[0080] The long short-term memory network model uses multiple hidden layers to learn and model time series data. When inputting data, the operation status data of the device is segmented by the sliding window method, the model is gradually trained, and based on the training results, the potential fault time window of the device is predicted.

[0081] The sliding window has a fixed length and slides over the time series data at a certain step size. Each time, the data within the window is intercepted as an input sample. For example, if the window length is 10 time steps and the step size is 1, the data from the 1st to the 10th time steps, the 2nd to the 11th time steps, etc. are intercepted in sequence as samples. This can convert the continuous time series data into multiple independent input samples for model training.

[0082] In this embodiment, the device operation status data at the current moment is input into the trained LSTM model. The model evaluates the health status of the device in real time according to the learned patterns and rules. By processing and calculating the input data, the current health score of the device is output.

[0083] Based on the historical operation status data and the current operation status data of the device, as well as the trained LSTM model, predict the possible types of faults that the device may occur. The model will identify the signals related to faults according to the characteristic patterns and change trends in the data and give the possible fault categories. Combining the prediction results of the model and time series analysis, determine the time window in which a fault may occur. Specifically, it is achieved by monitoring and analyzing the dynamic changes in the device operation status data and modeling the evolution of the fault occurrence probability over time. For example, predict that the device may have a certain fault within the next 1 - 3 days.

[0084] Integrate the possible types of faults that the device may occur and the corresponding time window information to generate the operation fault prediction result. This result is presented in the form of a report, clearly listing the predicted fault conditions and the possible time range. Provide the operation fault prediction result to the subsequent inspection and control module, and corresponding measures can be taken in a timely manner, such as arranging inspection tasks, preparing maintenance resources, etc.

[0085] 4. Inspection and Control Module

[0086] The inspection and control module is used to comprehensively judge the maintenance requirements of the device based on the device health status, operation fault prediction result, appearance fault prediction result, and environmental data, schedule and execute inspection tasks. The inspection tasks include ground equipment inspection and power line inspection. The ground equipment inspection is performed by robots, and the power line inspection is performed by drones. The inspection tasks are scheduled according to priorities, and the inspection results are fed back in real time after the inspection tasks are executed. Corresponding control operations are performed on the device according to the feedback results.

[0087] The inspection and control module manages tasks through the cloud scheduling system, specifically including:

[0088] Based on the device health status and the results of operation fault prediction, determine whether the device needs routine inspection. If the device health level is low or the fault prediction results indicate that a fault is about to occur, prioritize scheduling a robot or drone to perform inspection on this device;

[0089] Based on the results of appearance fault prediction, determine whether there are external visible faults or potential hidden dangers in the device. If the appearance fault prediction results show that the device has obvious appearance faults or hidden dangers, prioritize scheduling a robot or drone to perform inspection or repair operations on this device;

[0090] Comprehensively consider the device health status, the results of operation fault prediction and the results of appearance fault prediction, and dynamically adjust the priorities of each inspection task. The formula is:

[0091] P inspect =α·P health +β·P fault +γ·P appearance

[0092] Where, P inspect is the priority of the inspection task, P health is the priority of the device health status, P fault is the priority of the fault prediction results, P appearance is the priority of the appearance fault prediction results, and α, β, γ are weight coefficients;

[0093] During the task assignment process, according to the priority of the device inspection task and the geographical distribution, optimize the task assignment strategy through reinforcement learning to ensure that high-priority devices are inspected first.

[0094] After the inspection task is executed, the drone or robot records the device status detected during the inspection through image acquisition and sensor data, and uploads it to the cloud platform in real time. Then, it is processed by the intelligent analysis module to generate an inspection report, and based on the device historical fault data and predictive analysis, automatically generate a subsequent inspection plan and give a recommended solution for the next round of inspection; the inspection and control module can also trigger automatic repair or adjust the operating state of power equipment in case of an emergency fault; the intelligent analysis module calculates the health score using the aforementioned health score calculation formula.

[0095] Robots and drones use autonomous navigation and obstacle avoidance methods to perform inspection tasks; autonomous navigation is based on the fusion of lidar and vision sensors for positioning and path planning; during the inspection process, robots and drones analyze the device status by combining real-time acquisition of device operation data and environmental information with a deep learning model and determine whether it is necessary to perform emergency repairs or adjust the device operating state; during the inspection process, robots and drones perform path planning according to the following formula:

[0096]

[0097] Among them, Path(t) is the optimal path at the current moment t, and w i is the weight coefficient of each node in the path planning process. Cost(p i ) is the cost function of the path node p i . Calculating the cost of a node includes factors such as distance, obstacle avoidance, and time. This formula optimizes the path planning to ensure the efficient execution of the inspection task and effectively avoid collisions with obstacles.

[0098] The inspection and control module comprehensively analyzes the equipment health status, operation fault prediction results, appearance fault prediction results, and environmental data through a multi-dimensional data fusion method; the multi-dimensional data fusion method includes the Kalman filter and particle filter algorithms; the Kalman filter is used to smooth the time-series data of the equipment; the particle filter is applied to non-linear systems, and through simulating the trajectories of multiple particles, state prediction is carried out and the equipment health assessment results are optimized; the state estimation formula combining the Kalman filter and the particle filter is as follows:

[0099] x k = A·x k-1 + B·u k + w k

[0100] z k = H·x k + v k

[0101] Among them, x k is the system state, A is the state transition matrix, B is the control input matrix, u k is the control input, w k is the process noise, z k is the observed value, H is the observation matrix, v k is the measurement noise. Through the combined application of the Kalman filter and the particle filter, the system can achieve more accurate fault prediction and equipment health assessment in a multi-source data environment.

[0102] 5. Data Security and Transparency Module

[0103] The data security and transparency module is used to encrypt and store all the data in the system through the blockchain, and can generate a data query report for authorized users to query. The data query report includes the equipment operation status, historical fault records, and inspection task execution records.

[0104] Each data block in the blockchain network is verified through digital signatures. The encryption storage mechanism of the blockchain uses the SHA256 hash algorithm to perform hash processing on each data block, and the hash value is calculated by the following formula:

[0105] H(block) = SHA256(PreviousBlockHash || Data || Timestamp)

[0106] Among them, H(block) is the hash value of the current block, PreviousBlockHash is the hash value of the previous block, Data is the stored data of the current block, and Timestamp is the timestamp of the current block, ensuring that the hash value of each block is unique and closely associated with the data of the previous block, thus guaranteeing the immutability and chain integrity of blockchain data.

[0107] The blockchain is based on the public-private key encryption mechanism and uses asymmetric encryption methods to ensure that device data cannot be accessed or tampered with by unauthorized third parties during the transmission and storage processes; the public key is used to verify the authenticity of the data, and the private key is used for the encrypted storage of the data. All blockchain data is irreversibly encrypted and stored in the blockchain, and only authorized users can access and view it.

[0108] The data security and transparency module performs data access control through a permission management system, enabling users at different levels to access corresponding data and functions according to their authorized permissions; the permission management system includes a user identity authentication module and a permission allocation module. The user identity authentication module verifies users through multi-factor authentication methods, and the permission allocation module assigns different data access permissions according to the roles and responsibilities of different users;

[0109] The permission management system performs permission verification according to the following rules:

[0110] P(u) = {p1, p2,..., p n}

[0111] Among them, P(u) is the permission set of user u, and p1, p2,..., p n are the various permissions of this user, ensuring that each user can only access the data and function modules that they are authorized to.

[0112] The above is only the preferred embodiment of the present invention and is not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A power AI intelligent auxiliary control and inspection system, characterized in that, Including: A data acquisition module, which is used to collect the operation status data and appearance image data of power equipment through current sensors, voltage sensors, temperature sensors, humidity sensors, infrared thermal imaging cameras, and visual image sensors, and perform preprocessing; An image processing and analysis module, which is used to perform image processing based on the convolutional neural network model according to the appearance image data, identify whether there are appearance faults or potential hidden dangers in the equipment appearance, and generate appearance fault prediction results; A health assessment and fault prediction module, which is used to perform real-time assessment of the equipment health status based on the long short-term memory network model according to the operation status data, obtain the equipment health status, and predict possible faults and their time windows according to the equipment historical operation status and the equipment current operation status, and generate operation fault prediction results; An inspection and control module, which is used to comprehensively judge the maintenance requirements of the equipment according to the equipment health status, operation fault prediction results, appearance fault prediction results, and environmental data, schedule and execute inspection tasks. The inspection tasks include ground equipment inspection and power line inspection. The ground equipment inspection is executed by robots, and the power line inspection is executed by drones. The inspection tasks are scheduled according to priorities, and the inspection results are fed back in real time after the inspection tasks are executed. Corresponding control operations are performed on the equipment according to the feedback results; A data security and transparency module, which is used to encrypt and store all data in the system through blockchain, and can generate data query reports for authorized users to query. The data query reports include equipment operation status, historical fault records, and inspection task execution records.

2. The power AI intelligent auxiliary control and inspection system according to claim 1, characterized in that The long short-term memory network model uses multiple hidden layers to learn and model time series data. When inputting data, the operation status data of the equipment is segmented by the sliding window method, the model is gradually trained, and based on the training results, the potential fault time window of the equipment is predicted.

3. The power AI intelligent auxiliary control and inspection system according to claim 1, wherein The convolutional neural network model uses the appearance image data to extract the temperature distribution characteristics of each part of the equipment surface through multiple convolutional kernels, and judge whether there are appearance faults or potential hidden dangers; the convolutional neural network model includes a convolutional layer, a pooling layer, and a fully connected layer. The specific judgment process includes: In the first stage, feature extraction is performed on the input appearance image data through the convolutional layer to obtain the temperature change area in the equipment image; In the second stage, downsampling is performed on the output of the convolutional layer through the pooling layer to extract high-order features; In the third stage, high-order feature fusion is performed through the fully connected layer, and the classification result is output and it is judged whether there are appearance faults or potential hidden dangers in the equipment.

4. The power AI intelligent auxiliary control and inspection system according to claim 1, characterized in that The inspection and control module performs task management through a cloud scheduling system, specifically including: According to the equipment health status and operation fault prediction results, judge whether the equipment needs to be routinely inspected. If the equipment health degree is low or the fault prediction result indicates that a fault will occur soon, robots or drones are preferentially scheduled to inspect the equipment; According to the appearance fault prediction results, judge whether there are external visible faults or potential hidden dangers in the equipment. If the appearance fault prediction results show that the equipment has obvious appearance faults or hidden dangers, robots or drones are preferentially scheduled to check or repair the equipment; Based on the comprehensive device health status, operation fault prediction results, and appearance fault prediction results, dynamically adjust the priorities of each inspection task; During the task allocation process, according to the inspection task priorities and geographical distribution of the devices, optimize the task allocation strategy through reinforcement learning to ensure that high-priority devices are inspected first.

5. The power AI intelligent auxiliary control and inspection system according to claim 4, characterized in that, After the inspection task is executed, the drone or robot records the device status detected during the inspection through image acquisition and sensor data, and uploads it to the cloud platform in real time. Then, it is processed by the intelligent analysis module to generate an inspection report, and based on the device historical fault data and predictive analysis, automatically generate the subsequent inspection plan and give the recommended solution for the next round of inspection; The inspection and control module can also trigger automatic repair or adjust the operating state of the power equipment in case of an emergency fault.

6. The power AI intelligent auxiliary control and inspection system according to claim 1, characterized in that Each data block in the blockchain network is verified through digital signatures, and the encrypted storage mechanism of the blockchain uses the SHA256 hash algorithm to perform hash processing on each data block.

7. The power AI intelligent auxiliary control and inspection system according to claim 6, characterized in that, The blockchain is based on the public-private key encryption mechanism and uses asymmetric encryption methods to ensure that the device data will not be accessed or tampered with by unauthorized third parties during the transmission and storage processes; the public key is used to verify the authenticity of the data, and the private key is used for the encrypted storage of the data. All blockchain data is irreversibly encrypted and stored in the blockchain, and only authorized users can access and view it.

8. The power AI intelligent auxiliary control and inspection system according to claim 1, wherein The data security and transparency module performs data access control through the permission management system, enabling users at different levels to access corresponding data and functions according to the authorized permissions; the permission management system includes a user identity authentication module and a permission allocation module. The user identity authentication module verifies users through multi-factor authentication methods, and the permission allocation module assigns different data access permissions according to the roles and responsibilities of different users.

9. The power AI intelligent auxiliary control and inspection system according to claim 1, wherein The robot and the drone use autonomous navigation and obstacle avoidance methods to perform inspection tasks; the autonomous navigation is based on the fusion of lidar and visual sensors for positioning and path planning; during the inspection process, the robot and the drone analyze the device status by combining the real-time obtained device operation data and environmental information, and determine whether it is necessary to perform emergency repairs or adjust the device operating state.

10. The power AI intelligent auxiliary control and inspection system according to claim 1, characterized in that, The inspection and control module comprehensively analyzes the device health status, operation fault prediction results, appearance fault prediction results, and environmental data through a multi-dimensional data fusion method; the multi-dimensional data fusion method includes the Kalman filter and particle filter algorithms; the Kalman filter is used to smooth the time-series data of the device; the particle filter is applied to nonlinear systems, and through simulating multiple particle trajectories, performs state prediction and optimizes the device health assessment results.

Images