Research On Intelligent Identification Of Violations In Substation Reconstruction And Expansion Based On Artificial Intelligence

In recent years,with the rapid development of my country’s national economy and the continuous improvement of people’s living standards,some old substations can no longer meet the needs of people’s production and life.It is imperative to renovate and expand the old substation.However,the reconstruction and expansion of the old substation is different from the conventional civil construction construction.There are many electrical equipment and power cables inside the substation,and the substation is not allowed to stop running during the reconstruction and expansion construction,otherwise it will not be able to supply power normally,which wi ll eventually cause heavy damage to people’s property.Therefore,in the process of reconstruction and expansion of old substations,only a local construction scheme of small area power failure can be adopted,which makes the construction area adjacent to live facilities become a high-risk area,which leads to the safety management and control of substation reconstruction and expansion construction is also different from other constructions scenarios.Specifically,in the construction scene of substation reconstruction and expansion,in addition to the conventional safety helmet wearing detection,personnel gathering detection is also particularly important.The reason is that the substation reconstruction and expansion construction has strict construction specifications,and it is clearly required that people in the construction area shall not gather.On the other hand,in order to prevent the danger of electric shock during the construction process,the operation location should be separated from the live facilities by a safe distance.The traditional method usually uses the manual construction of physical fences to divide the safety area.However,this method is time-consuming and labor-intensive and cannot achieve intelligent early warning.More importantly,due to the elevation limit of the physical fence,it is impossible to detect the high-altitude violations of construction personnel and equipment.Therefore,it will bring great safety hazard to the reconstruction and expansion of the substation.YOLOv5(You Only Look Once)is a target detection and location algorithm based on deep neural network.Its running speed is fast and its identification accuracy is high.It can meet the requirements of real-time detection in substation reconstruction and expansion scenarios,but its positioning accuracy is insufficient.In response to the above problems,this thesis proposes a conventional illegal behavior recognition algorithm and a cross-border illegal behavior detection algorithm to detects object such as helmets based on the YOLOv5 algorithm,combined with RTK(Real-time kinematic)positioning technology,to make up for the problem of insufficient visual positioning accuracy,and deployed in the Android platform,aiming to solve the problem of real-time security management and control in substations.The main research contents and innovations of this thesis are as follows:1.This thesis proposes a detection algorithm for conventional illegal behavior in substation renovation and expansion.For the problem that the safety helmet in substation is small and difficult to detect,the attention module CAM(Channel Attention Module)is added to YOLOv5 to improve the performance of network detection on the safety helmet.On this basis,a discriminant algorithm is proposed to judge the clustering behavior of workers2.This thesis proposes a detection algorithm for cross-border illegal behavior in substation renovation and expansion.To solve the problem that the triangular cones in the substation are small and difficult to detect,the attention module CBAM(Convolutional Block Attention Module)is added to YOLOv5 to make the important object features occupy a larger proportion of network processing and enhance the feature learning ability of the network to the object region.Alpha-iou was selected as the loss function of bounding box regression to enhance the robustness of small data sets.And proposed a ground perimeter out-of-bounds discrimination algorithm based on PNPOLY,a high-altitude perimeter out-of-bounds discrimination algorithm based on high-altitude projection,and an elevation out-of-bounds discrimination algorithm based on visual RTK integration,to judge the safety status of operators in real time.3.The object detection algorithm and violation detection method proposed in this thesis are deployed on the mobile terminal,and the safety detection is applied in the substation scenario,so as to realize the detection of the wearing state of the operator’s helmet,the detection of whether workers gather and the detection of workers entering dangerous areas.