Research On Insulator Location Of Automatic Catenary Hot Washing

To avoid the flashover accidents of the electrified railway insulators,it is necessary to clean them regularly.So far,the most commonly used insulator cleaning tool is KJ series insulators water flushing vehicle,but the whole insulator washing process mainly depends on manual experience,which has many disadvantages such as difficult operation,high intensity and large hazard coefficient.Therefore,it is of great significance to study the automatic and intelligent insulators flushing equipment to maintain the electrified railway.Some scholars have already done some research based on machine vision and image recognition theory,but so far,there is no research on automatic flushing equipment.In this paper,the control of KJ series water flushing vehicle was studied based on the reinforcement learning theory and the experience of water flushing experts,through the establishment of different insulator locating learning model,electrified railway insulator water flushing expert system was developed,by training and simulating various flushing mode,the expert system can improve the accuracy of automatic flushing,fast,efficient and intelligent,by effectively preventing the loss of the target and correcting the deviation.The following aspects were included in this article:First of all,the overall composition of the electric insulator water flushing control system for railway insulators was introduced,including the flushing water gun system and the flushing control system.The functional analysis of the water flushing control system was carried out,and the operation flow was divided into three parts: the flushing preparation phase,the flushing phase and the after-flushing phase.The overall working flow of the automatic electrified water flushing system was described in detail.Then different corresponding reinforcement learning model is proposed for different flushing environment.A railway catenary insulator positioning and tracking method is proposed in preset the flushing position stage based on heuristic Q(?)study,combined with the sensor information,the railway insulator water flushing environment model was established,by searching the pillar feature and adjusting the reward function,heuristic strategy function was designed and integrated into Q(?)learning to improve the learning speed and the convergence rate of the algorithm.During single cannon flushing process,the railway insulators water flushing and positioning method was studied based on the least squares support vector machine(LS-SVM)and Q(?)learning.The flushing behavior of water cannon was discretized under continuous action space,LS-SVM was used to estimate the Q value online,by rolling time window to update training sample.The simulation results showed that the method can significantly improve the learning speed and convergence of single cannon flushing,and at the same time,by comparing with other insulator location methods,the speed of the method was obviously improved,which was more consistent with the actual needs of engineering.Besides,aiming at the double canon cooperated model and the double canon separated model,the option hierarchical multi-agent reinforcement learning method was proposed,by introducing the concept of SMDP,by constructing the options on state space and executing the corresponding strategy to decreased the learning dimensions,while sharing the experience and the knowledge of intelligent through the blackboard model,as well as the local agent,global agent two layer model,which made the learning effect meet with the actual engineering requirements.Finally,based on the Clips,VB language and the matlab simulation mode,the insulator water flushing expert system was developed.The basic structure of the expert system was introduced,including the knowledge base,the reasoning mechanism,the main function and so on.In the technology side,the main structure was designed by using Clips,including both the knowledge base and reasoning mechanism,then water flushing system interface was developed by using programming friendly VB language,including user login system,expert system and simulation module,at the same time the system can realize the offline training procedure of electrified railway insulator cleaning process simulation by calling Matlab code.At last,both the different outside flushing condition、different flushing mode、different insulator distribution and different vichel speed were tested and analyzed,the results showed that the automatic flushing expert system designed in this paper accords with the actual engineering requirements of the catenary electrified water flushing operation,and provides a new solution for the software part of the automatic water equipment.