Optimized Neural Network by Genetic Algorithm and Its Application in Fault Diagnosis of Three-level Inverter

Abstract

Multilevel inverters have been widely applied in high-voltage and high-power applications. Therefore, fault diagnosis of such circuits is becoming more and more important. Fault diagnosis for single device open-circuit fault of three-level inverter based on BP (back propagation) neural network is studied in this paper. One of the weak-points of BP algorithm which is commonly used is that the optimal procedure is easily stacked into the local minimal value and cause strict demands of initial value. So a fault diagnosis method based on BP neural network and genetic algorithm (GA) is proposed in this paper. Firstly, bridge voltage of three-level inverter is collected as fault signal and feature is extracted to determine the structure of the BP neural network. After this, GA is applied to optimize the initial weights and thresholds of BP neural network, and then the network is trained to diagnose faults of three-level inverter to determine the specific failure device. The simulation result shows that the method can isolate fault modes proposed exactly, and the weak-point of network can effectively avoid, improve the diagnostic accuracy.