Accurate Reading And Reliability Estimation Method For Substation Meter

Nowadays,with the rapid development of the national economy,society’s demand for electricity is growing,the security of power systems is becoming more and more important.As the core component of the power system,substation is becoming intelligent in unmanned inspection,the robot is gradually replacing the manual to complete the inspection of the substation equipment.Robot’s inspection mainly includes meter reading and temperature measurement.For the problem that start and end tick marks of the meter are not accurate enough,the traditional Hough line pointer recognition method is not accurate enough and there is no reliability estimation model for existing meter readings,this thesis proposes a more accurate meter reading method and corresponding reliability estimation model;in addition,for the problem that the image quality of the meter is degraded in the rainy environment and the robot reading is difficult,this thesis also proposes a multi-detail residual network de-raining model,which better solves the rainy environment accurate reading of the meter.The main research contents and contributions of this paper are as follows:(1)For the problem that the start and the end of the meter tick position extraction is not accurate enough,the standard Hough line improvement algorithm is easily affected by the parameters to generate pointer positioning deviation,this thesis proposes an adaptive instrument reading algorithm.By obtaining the circle of the tick mark through the Hough ring detection algorithm,and then use the threshold judgement to extract the start and the end positions of the tick mark;combining the ring of the tick mark,and further extract the tip of the pointer by reducing the radius of the ring,thus the thesis can obtain a kind of meter reading method.The experimental results show that the accuracy of the proposed algorithm increases by an average of 4.5%,and it is stable and reliable.(2)Considering the existing meter reading algorithms need a large number of parameters,there will be a large reading deviation when the parameter adjustment is not correct,and there is no model available to judge the reliability of the reading right now.This thesis proposes a reading reliability judgment model.The model uses the clustering idea to realize the consistent and significant association constraint detection of the instrument features,which is used to determine the position of the pointer center of the instrument.Then the 3D Gaussian hybrid reliability model is constructed to evaluate whether the pointer recognition result meets the reading standard,realizing the reliability estimation of the meter reading algorithm.(3)Considering the problem that the image noise increased and the sharpness is reduced by the robot inspection due to the rainy weather,which makes the subsequent reading difficult,a multi-detail residual network model is proposed.This network model uses guided filtering to decompose the image in order to obtain the smooth image and detailed images of different frequencies.The multi-detail residual network is used to learn the residual mapping relationship between rain image and rainless image,which removes the interference of rainwater,solves the problem of difficult reading of pointer meter in rainy environment,and lays a foundation for the realization of all-weather inspection of substation robot.