Research On Distributed Identification Algorithm And Unloading Strategy For Transmission Line Inspection

The safe and stable operation of transmission line equipment is the key to the normal operation of the power grid system.It is of great significance for the safe operation of the power grid to adopt accurate and efficient detection methods to detect the faults of transmission line equipment in time.With the advancement of smart grid construction,transmission line inspections have changed from traditional manual inspections to intelligent inspections.However,the long inspection period and the cumbersome inspection process are still the problems that need to be solved urgently in the inspection of transmission lines.Therefore,this paper takes the insulator in the transmission line equipment as the research object,and proposes a distributed identification algorithm and its unloading strategy for transmission line inspection to improve the detection efficiency.Firstly,in view of the problems that the existing power transmission and transformation equipment detection algorithm model is too large,time-consuming and cannot be deployed at the edge,an identification method of lightweight insulators at the edge is proposed.The YOLOv4 target detection algorithm in the one-stage target detection is selected,and the lightweight Mobile Netv3 network is used to replace the CSPDarknet53 feature extraction network in YOLOv4 to reduce the model network parameters.In order to improve the accuracy of the recognition algorithm,the activation function of the Mobile Netv3 network and the loss function of the YOLOv4 algorithm are respectively improved.For the Mobile Netv3 network,PRe LU is used instead of Re LU to become the activation function of the shallow part of the network.For the loss function of the YOLOv4 target detection algorithm,the Focal loss function is used to increase the weight of the target sample,so that the model can focus on the detection of the target object and speed up the regression of the prediction frame.It reduces the interference of the unbalanced positive and negative samples on the model and improves the accuracy of model detection.The experimental results show that the lightweight insulator defect detection method proposed in this paper can greatly reduce the size of the model while ensuring the detection accuracy,and can be deployed at the edge.Secondly,for the light-weight transmission line equipment fault detection algorithm proposed in the previous chapter,a deep neural network partitioning algorithm based on binary particle swarm is proposed,so that it can be reasonably deployed on Unmanned Aerial Vehicle(UAV)and edge servers.First,analyze the impact of different network layer division points of Mobile Netv3 improved in the previous chapter on the overall transmission line inspection system,then establish the system energy consumption and delay model,and formulate the total system cost.The problem of minimizing the total system cost is transformed into a binary particle swarm position optimization problem.In order to improve the speed of obtaining the optimal partition point,the chaotic map is used to optimize the initial position of the particle.Secondly,the mutation strategy is adopted to avoid the stagnation of particle search and further improve the performance of the partition algorithm.The experimental results show that the deep neural network partition algorithm proposed in this paper realizes the reasonable division of the deep neural network on the UAV and the edge server,and greatly improves the efficiency of insulator self-explosion fault detection.Finally,it is aimed at the problem of coordinated computing and unloading of multiple UAVs for simultaneous inspection of transmission line equipment,and the problem of allocating computing resources for multi-UAV inspection of transmission lines.A multi-UAV cooperative offloading strategy based on deep reinforcement learning is proposed.The multi-UAV cooperative unloading problem in transmission line inspection is formulated as a Markov game,and then a deep deterministic policy gradient algorithm is used to solve the problem.In addition,the association between UAVs is considered to increase the mutual learning between UAVs.The experimental simulation results show that,compared with other unloading strategies,the proposed multi-UAV cooperative unloading strategy can effectively improve the inspection efficiency of transmission lines.