Power plant safety monitoring system based on artificial intelligence
By adopting a power plant safety monitoring system based on artificial intelligence in the power plant, using multi-dimensional data extraction and semi-supervised learning algorithm to monitor equipment and personnel, the problem of unmonitored old equipment and difficult to analyze big data is solved, and efficient power plants are achieved Safety monitoring and failure prediction.
Patent Information
- Application Number
- CN202411180725.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-27
- Publication Date
- 2024-11-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Old equipment in the power plant is not integrated into modern monitoring systems, which may cause the early warning system to miss out on safety hazards, and the monitoring data is large, making it difficult to determine the authenticity of abnormal situations.
A power plant safety monitoring system based on artificial intelligence was designed, including power plant equipment monitoring module, power plant personnel monitoring module and safety reporting module. Using multi-dimensional data extraction, semi-supervised learning algorithms, convolutional neural network and other technologies, the equipment and Abnormalities for the personnel and generate solutions.
Accurate monitoring of power plant equipment and personnel, timely discover abnormal trends, predict potential failures, improve maintenance efficiency, and ensure the stability of power supply.
Smart Images

Figure CN118938770A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power plant safety monitoring, and in particular to a power plant safety monitoring system based on artificial intelligence. Background Art
[0002] At present, some old equipment in power plants may not have been integrated into modern monitoring systems, resulting in the early warning system missing safety hazards due to incomplete monitoring. In addition, for power plants, the volume of monitoring data is very large, and how to determine the authenticity of the safety anomalies obtained through decision-making needs to be urgently addressed. Summary of the invention
[0003] In order to solve the above technical problems, the present invention provides a power plant safety monitoring system based on artificial intelligence, the system comprising:
[0004] Power plant equipment monitoring module, power plant personnel monitoring module and safety reporting module;
[0005] The power plant equipment monitoring module is used to monitor whether the power plant facilities are operating normally. If there is any abnormal operation, the abnormal information is sent to the safety reporting module;
[0006] The power plant personnel monitoring module is used to monitor the personnel safety of the power plant personnel and send abnormal information to the safety reporting module if abnormal behavior of personnel is found;
[0007] The safety reporting module is used to receive abnormal information and process it separately to achieve safety monitoring of the power plant.
[0008] Optionally, the power plant equipment monitoring module includes a multi-dimensional data extraction submodule, an operation status monitoring submodule and an equipment abnormality information sending submodule;
[0009] The multi-dimensional data extraction submodule is used to collect multi-dimensional angle monitoring data of power plant equipment and obtain multi-dimensional monitoring data after preprocessing;
[0010] The operation status monitoring submodule is used to monitor the operation status of the equipment based on the multi-dimensional monitoring data using a semi-supervised learning algorithm to identify whether the equipment is abnormal;
[0011] The abnormal information sending submodule is used to receive an abnormal signal sent when a device is abnormal.
[0012] Optionally, the workflow of the multi-dimensional data extraction submodule specifically includes:
[0013] In the extraction of multi-dimensional data, multi-dimensional data of power plant equipment is obtained based on camera devices and sensors;
[0014] The process of using the camera device to obtain data of power plant equipment specifically includes:
[0015] Obtain images captured from multiple camera perspectives as basic device data;
[0016] Use temperature sensors, speed sensors, pressure sensors, flow sensors and differential expansion sensors to monitor power plant equipment and obtain multi-dimensional sensor data;
[0017] The multi-dimensional data of the sensor is integrated with the basic data of the equipment to obtain multi-dimensional data of the power plant equipment.
[0018] Optionally, the workflow of the operation status monitoring submodule specifically includes:
[0019] Using semi-supervised learning to perform quality assessment on the multi-dimensional data of the power plant equipment and generate pseudo-label images;
[0020] Preprocessing the pseudo-label image to obtain a preprocessed image, and inputting the preprocessed image into an aggregated attention mechanism to extract global features and local features of multi-dimensional data;
[0021] The global features and the local features are input into an attention fusion mechanism for integration to obtain an integrated image, and abnormal monitoring of the operating status of the device is achieved based on the integrated image.
[0022] Optionally, the power plant personnel monitoring module includes: an employee trajectory safety identification submodule, an employee behavior safety identification submodule and an employee abnormal information sending submodule;
[0023] The employee trajectory safety identification submodule is used to monitor the employee trajectory in conjunction with the monitoring results of the power plant equipment monitoring module;
[0024] The employee behavior safety identification submodule is used to use a convolutional neural network to identify unsafe behaviors of employees and send abnormal employee behavior information;
[0025] The employee abnormality information invention submodule is used to receive and send employee behavior abnormality information.
[0026] Optionally, the workflow of updating the dangerous area of the employee trajectory safety identification module in conjunction with the monitoring result of the power plant equipment monitoring module in the employee trajectory safety identification module specifically includes:
[0027] Based on the abnormal monitoring results of power plant equipment, the dangerous area around the power plant equipment is divided. According to the employee trajectory monitoring results, deep inverse reinforcement learning is used to predict the employee trajectory. When the predicted trajectory is within the dangerous area, the trajectory abnormality information is sent.
[0028] Optionally, the workflow of the employee behavior safety identification submodule specifically includes:
[0029] Using a U-Net network to extract the appearance features and motion features of employees in the employee behavior video frame, and reconstructing the next frame image based on the appearance features and the motion features;
[0030] Given a preset number of employee behavior video frames, perform spatiotemporal positioning to obtain a spatiotemporal positioning map, build a behavior prediction model based on the VideoCapsuleNet architecture, input the spatiotemporal positioning map into the behavior prediction model training, and obtain a trained behavior prediction model;
[0031] The reconstructed graph is input into the behavior detection model for prediction, and the predicted behavior is compared with the abnormal behavior to achieve safe identification of employees.
[0032] Optionally, the security reporting module includes an abnormal information summary submodule, a solution generation submodule and a communication submodule;
[0033] The abnormal information summary submodule is used to receive abnormal information sent by the power plant equipment monitoring module and the power plant personnel monitoring module;
[0034] The solution generation submodule is used to generate a solution based on the abnormal information;
[0035] The communication submodule is used to send the abnormality information and the solution to the technician's terminal at the same time.
[0036] Compared with the prior art, the present invention has the following beneficial effects:
[0037] Based on the analysis results of monitoring data, the present invention can detect abnormal trends in real time and predict potential faults, thus ensuring the stability of power supply. Based on multiple algorithms of artificial intelligence, the present invention can accurately identify the type of fault and improve maintenance efficiency. The present invention monitors both equipment and employees, associates the trajectory of people with abnormal equipment information, and avoids abnormal situations based on abnormal equipment. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0039] Figure 1 4 is a system structure diagram of a safety monitoring system according to an embodiment of the present invention. DETAILED DESCRIPTION
[0040] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0041] Example
[0042] An artificial intelligence-based power plant safety monitoring system, such as Figure 1 As shown, the system includes: a power plant equipment monitoring module, a power plant personnel monitoring module and a safety reporting module;
[0043] The power plant equipment monitoring module is used to monitor whether the power plant facilities are operating normally. If an abnormal operation occurs, the abnormal information is sent to the safety reporting module. The power plant equipment monitoring module includes a multi-dimensional data extraction submodule, an operation status monitoring submodule and an equipment abnormal information sending submodule; the multi-dimensional data extraction submodule is used to collect multi-dimensional angle monitoring data of the power plant equipment, and obtain multi-dimensional monitoring data after preprocessing; the operation status monitoring submodule is used to monitor the equipment operation status based on the multi-dimensional monitoring data using a semi-supervised learning algorithm to identify whether the equipment is abnormal; the abnormal information sending submodule is used to receive abnormal signals sent when the equipment is abnormal.
[0044] The workflow of the multi-dimensional data extraction submodule specifically includes: in the extraction of multi-dimensional data, multi-dimensional data of power plant equipment is obtained based on a camera device and a sensor; the process of using the camera device to obtain data of power plant equipment specifically includes: obtaining images captured from multiple camera perspectives as basic equipment data; using a temperature sensor, a speed sensor, a pressure sensor, a flow sensor and a differential expansion sensor to monitor data of the power plant equipment to obtain sensor multi-dimensional data; and fusing the sensor multi-dimensional data with the equipment basic data to obtain multi-dimensional data of the power plant equipment.
[0045] The working process of the operation status monitoring submodule specifically includes: using semi-supervised learning to perform quality assessment on the multi-dimensional data of the power plant equipment to generate a pseudo-label image; preprocessing the pseudo-label image to obtain a preprocessed image, and inputting the preprocessed image into an aggregated attention concentration mechanism to extract global features and local features of the multi-dimensional data; inputting the global features and the local features into an attention fusion mechanism for integration to obtain an integrated image, and realizing abnormal monitoring of the equipment operation status based on the integrated image.
[0046] Specifically, it includes: inputting multi-dimensional data into the semi-supervised recovery model to generate pseudo-label images, using URanker as the pseudo-reference truth value to evaluate the pseudo-label images; according to the evaluation results, selecting pseudo-label images to be stored in the Reliable Bank. Among the images acquired from the power plant, some images are subjected to the most absorption and scattering of light due to various reasons, resulting in hue deviation of the images, which affects subsequent image analysis. The pseudo-label images are color corrected to obtain preprocessed images; the process of color correcting the pseudo-label images to obtain preprocessed images specifically includes: color correcting the pseudo-label images stored in the Reliable Bank through local adaptive color correction. The aggregated attention mechanism can use the preprocessed image as the feature map L1, and use convolution kernels of sizes 1, 3, and 5 and a mean pooling layer to extract feature information, and splice the obtained four feature maps; use a 1×1 convolution kernel to change the number of channels of the spliced feature map and add it to the input feature map L1 to obtain L2:
[0047] L2=conv1(concat(conv1(pool(L1),conv1(L1),conv3(L1),conv5(L1)))+L1
[0048] Copy L2, activate the first copy using the symmetric sigmoid function, use the result of the activation as the weight, and multiply it with the other copy of L2 to get L3:
[0049] L3=sigmoid +0.5 (L2)×L2
[0050] Normalize L3 layer by layer, and then calculate the residual to get L4:
[0051] L4=LayerNorm(L3)+L3
[0052] L4 is input into the multi-layer perception mechanism for layer normalization, and the residual is calculated with L4 to obtain L5:
[0053] L5 = LayerNorm (MLP (L4)) + L4;
[0054] Based on L1, L2, L3, L4 and L5, an attention fusion mechanism is used to extract global and local information.
[0055] The global features and the local features are input into an attention fusion mechanism for integration to obtain an integrated image, and abnormal monitoring of the operating status of the device is achieved based on the integrated image.
[0056] The power plant personnel monitoring module is used to monitor the personnel safety of the power plant personnel. If any abnormal behavior of the personnel is found, the abnormal information is sent to the safety reporting module.
[0057] The power plant personnel monitoring module includes: an employee trajectory safety identification submodule, an employee behavior safety identification submodule and an employee abnormal information sending submodule; the employee trajectory safety identification submodule is used to monitor the employee trajectory in conjunction with the monitoring results of the power plant equipment monitoring module; the employee behavior safety identification submodule is used to use a convolutional neural network to identify unsafe behaviors of employees and send employee behavior abnormal information; the employee abnormal information invention submodule is used to receive and send employee behavior abnormal information.
[0058] The workflow of updating the dangerous area of the employee trajectory safety identification module in combination with the monitoring results of the power plant equipment monitoring module in the employee trajectory safety identification module specifically includes: dividing the dangerous area around the power plant equipment based on the abnormal monitoring results of the power plant equipment, and using deep inverse reinforcement learning to predict the employee trajectory according to the employee trajectory monitoring results. When the predicted trajectory is within the dangerous area, the trajectory abnormality information is sent.
[0059] The improved policy generation subnet and the improved trajectory fusion subnet are fused to obtain the multimodal employee trajectory prediction results. The dataset preprocessing part aims to extract the existing real dataset into raster images and historical sequence information used by the network. Specifically, the static scene layout information is extracted as raster images in the form of image rasterization, and the motion information of the dynamic scene is extracted as historical sequence information in the form of vectors to prepare for the trajectory prediction of the subsequent network model. The policy generation subnet is mainly based on inverse reinforcement learning. It learns a policy that is not constrained by the predefined target state based on the historical trajectory information and raster images of the surrounding vehicles, and explores the optimal policy in a two-dimensional grid. The trajectory fusion subnet first samples according to the optimal policy to generate paths to different reasonable targets in the two-dimensional grid. Each state sequence sampled from the policy is taken as a plan and used as the input of the next part. The recursive neural network encoding-decoding method with an attention mechanism and a loss function with a correction factor added to output a continuous-valued trajectory conditioned on the sampling plan are adopted.
[0060] The workflow of the employee behavior safety identification submodule specifically includes: using the U-Net network to extract the appearance features and motion features of employees in the employee behavior video frame, and reconstructing the next frame image based on the appearance features and the motion features; given a preset number of employee behavior video frames, performing spatiotemporal positioning to obtain a spatiotemporal positioning map, building a behavior prediction model based on the VideoCapsuleNet architecture, inputting the spatiotemporal positioning map into the behavior prediction model training, and obtaining a trained behavior prediction model; inputting the reconstructed map into the behavior detection model for prediction, and comparing the predicted behavior with the abnormal behavior to achieve safe identification of employees.
[0061] The safety reporting module is used to receive abnormal information and process it separately to realize safety monitoring of the power plant. The safety reporting module includes an abnormal information summary submodule, a solution generation submodule and a communication submodule; the abnormal information summary submodule is used to receive abnormal information sent by the power plant equipment monitoring module and the power plant personnel monitoring module; the solution generation submodule is used to generate a solution based on the abnormal information; the communication submodule is used to send the abnormal information and the solution to the technician's terminal at the same time.
[0062] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.
Claims
1. A power plant safety monitoring system based on artificial intelligence, It is characterized in that The system includes: a power plant equipment monitoring module, a power plant personnel monitoring module and a safety reporting module; The power plant equipment monitoring module is used to monitor whether the power plant facilities are operating normally. If there is any abnormal operation, the abnormal information is sent to the safety reporting module; The power plant personnel monitoring module is used to monitor the personnel safety of the power plant personnel and send abnormal information to the safety reporting module if abnormal behavior of personnel is found; The safety reporting module is used to receive abnormal information and process it separately to achieve safety monitoring of the power plant.
2. According to the artificial intelligence-based power plant safety monitoring system of claim 1, It is characterized in that The power plant equipment monitoring module includes a multi-dimensional data extraction submodule, an operation status monitoring submodule and an equipment abnormality information sending submodule; The multi-dimensional data extraction submodule is used to collect multi-dimensional angle monitoring data of power plant equipment and obtain multi-dimensional monitoring data after preprocessing; The operation status monitoring submodule is used to monitor the operation status of the equipment based on the multi-dimensional monitoring data using a semi-supervised learning algorithm to identify whether the equipment is abnormal; The abnormal information sending submodule is used to receive an abnormal signal sent when a device is abnormal.
3. According to the artificial intelligence-based power plant safety monitoring system of claim 2, It is characterized in that The workflow of the multi-dimensional data extraction submodule specifically includes: In the extraction of multi-dimensional data, multi-dimensional data of power plant equipment is obtained based on camera devices and sensors; The process of using the camera device to obtain data of power plant equipment specifically includes: Obtain images captured from multiple camera perspectives as basic device data; Use temperature sensors, speed sensors, pressure sensors, flow sensors and differential expansion sensors to monitor power plant equipment and obtain multi-dimensional sensor data; The multi-dimensional data of the sensor is integrated with the basic data of the equipment to obtain multi-dimensional data of the power plant equipment.
4. According to the artificial intelligence-based power plant safety monitoring system of claim 3, It is characterized in that The workflow of the operation status monitoring submodule specifically includes: Using semi-supervised learning to perform quality assessment on the multi-dimensional data of the power plant equipment and generate pseudo-label images; Preprocessing the pseudo-label image to obtain a preprocessed image, and inputting the preprocessed image into an aggregated attention mechanism to extract global features and local features of multi-dimensional data; The global features and the local features are input into an attention fusion mechanism for integration to obtain an integrated image, and abnormal monitoring of the operating status of the device is achieved based on the integrated image.
5. According to the artificial intelligence-based power plant safety monitoring system of claim 4, It is characterized in that The power plant personnel monitoring module includes: an employee trajectory safety identification submodule, an employee behavior safety identification submodule and an employee abnormal information sending submodule; The employee trajectory safety identification submodule is used to monitor the employee trajectory in conjunction with the monitoring results of the power plant equipment monitoring module; The employee behavior safety identification submodule is used to use a convolutional neural network to identify unsafe behaviors of employees and send abnormal employee behavior information; The employee abnormality information invention submodule is used to receive and send employee behavior abnormality information.
6. According to the artificial intelligence-based power plant safety monitoring system of claim 5, It is characterized in that The workflow of updating the dangerous area of the employee trajectory safety identification module in conjunction with the monitoring results of the power plant equipment monitoring module in the employee trajectory safety identification module specifically includes: Based on the abnormal monitoring results of power plant equipment, the dangerous area around the power plant equipment is divided. According to the employee trajectory monitoring results, deep inverse reinforcement learning is used to predict the employee trajectory. When the predicted trajectory is within the dangerous area, the trajectory abnormality information is sent.
7. According to claim 5, the power plant safety monitoring system based on artificial intelligence, It is characterized in that The workflow of the employee behavior safety identification submodule specifically includes: Using a U-Net network to extract the appearance features and motion features of employees in the employee behavior video frame, and reconstructing the next frame image based on the appearance features and the motion features; Given a preset number of employee behavior video frames, perform spatiotemporal positioning to obtain a spatiotemporal positioning map, build a behavior prediction model based on the VideoCapsuleNet architecture, input the spatiotemporal positioning map into the behavior prediction model training, and obtain a trained behavior prediction model; The reconstructed graph is input into the behavior detection model for prediction, and the predicted behavior is compared with the abnormal behavior to achieve safe identification of employees.
8. The power plant safety monitoring system based on artificial intelligence according to claim 1, It is characterized in that The security reporting module includes an abnormal information summary submodule, a solution generation submodule and a communication submodule; The abnormal information summary submodule is used to receive abnormal information sent by the power plant equipment monitoring module and the power plant personnel monitoring module; The solution generation submodule is used to generate a solution based on the abnormal information; The communication submodule is used to send the abnormality information and the solution to the technician's terminal at the same time.
