| There are all kinds of power equipment installed in the power system.In the daily operation,the operation state the power equipment is directly related to the security and stability of the power system.Some research shows that more than half of the faults of power equipment are relevant to the heating of power components,and as an important part carrying load current,the conductive joints of substation equipment are also the key parts that are prone to the hidden danger of overheating defects.Therefore,in order to solve this problem,an automatic joint heating recognition method of substation equipment based on Mask R-CNN and Support Vector Machine is proposed,which can automatically locate the conductive joint parts and judge the heating defects in substation equipment images taken by substation robots,so as to improve the efficiency and accuracy of automatic joint heating recognition of substation equipment.Firstly,the algorithm theory of Mask R-CNN model and Support Vector Machine model is introduced.The structures,characteristics and algorithms of the networks in the two prediction stages of Mask R-CNN model are analyzed,the algorithms and characteristics of Support Vector Machine model are studied,and the kernel function is selected.According to the characteristics of substation equipment joint recognition,Mask R-CNN model and Support Vector Machine model are designed.The model design of Mask R-CNN includes the improvement of network structure,the preprocessing of training images and the construction of the training method,and the model design of Support Vector Machine includes the analysis of model function,the construction of input vectors and the construction of output labels.Meanwhile,according to the precision requirement of the substation equipment conductive joint recognition task,the hyperparameters of Mask R-CNN and Support Vector Machine are set.Finally,the heating recognition algorithm of substation equipment joints based on Mask R-CNN and Support Vector Machine is put forward.On the basis of establishing the flow of the substation equipment joint heating recognition algorithm,a joint position prediction algorithm based on Mask R-CNN,a joint position judgment algorithm based on Support Vector Machine and a joint heating judgment algorithm for substation equipment are proposed,and 1000 groups of substation equipment images taken by substation robots are adopted to carry out an experimental study on the joint heating recognition algorithm of substation equipment.The experimental results show that the accuracy of the proposed joint heating defect recognition model for substation equipment images can reach 98.8%,which has obvious advantages and thus can effectively improve the accuracy of joint heating defect recognition of substation equipment. |
