In view of such problems of substation inspection robot as difficulty in planning smooth route under tradi-tional moving plan method and unpredictable action under uncertain condition

the study of deeply reinforced learn-ing method of the moving planning issue of substation inspection robot under uncertain condition is proposed. In thispaper

the influence of rewarding value model II/exploring strategies and neural network structure in the deep learn-ing on the entire moving plan is analyzed. The different neural network with different structure is designed and the rel-evant comparison experiment is performed. It is indicated by the result that in the current task scenario

the calcula-tion time of C2 neural network structure is shorter than the C1 neural network structure and C3 neural network struc-ture in the same operand.Therefore

in case of shortage of computation resource

the suggestion of adopting C2 neuralnetwork structure is more beneficial for accurate moving plan of substation inspection robot and also the improvementof automatic controlling efficiency.