Abstract: An optimal sensor deployment method based on hybrid information entropy index is proposed to improve the fault diagnosis performance of power supply vehicles (PSVs). Low fault diagnosability of PSVs is one of the main causes of their high failure rate, and the unbalanced configuration of the measuring point sensors is the key to the difficulty of rapid and reliable detection of power vehicle faults. Thus, we can first obtain the posterior probability of sensor residual after fault using the Bayes theorem, and then calculate the Value of information (VOL) of the posterior probability for PSV fault diagnosis. Second, due to the possibility of redundancy between sensors, it is difficult for sensors to interpret the numbers and locations simultaneously. Therefore, the redundancy between sensors is further quantified by using the Transfer Entropy (TE) method, and the optimal sensor configuration problem is reduced to a multi-objective optimization problem of solving VOL and TE, which reveals the best sensor configuration. Simulation results show that the hybrid information entropy method has obvious advantages for clarifying and analyzing the maximum coverage space and the minimum configuration set of sensors in the process of fault diagnosis.

Key words: valueofinformation, transferentropy, faultdiagnosability, sensorplacement, powersupplyvehicle