Fault Diagnosis and Reconfiguration of Multilevel Inverter Switch Failure-A Performance Perspective

Abstract

Multilevel Inverters (MLI) gains importance in Distribution systems, Electrical Drive systems, HVDC systems and many more applications. As Multilevel Inverters comprises of number of power switches the fault diagnosis of MLI becomes tedious. This paper is an attempt to develop and analyze the fault diagnosis method that utilizes Artificial Neural Network to get it trained with the fault situations. A performance analysis of Genetic Algorithm (GA) and the Modified Genetic Algorithm (MGA), which optimizes the Artificial Neural Network (ANN) that trains itself on the fault detection, and reconfiguration of the Cascaded Multilevel Inverters (CMLI) is attempted. The Total Harmonic Distortion (THD) occurring due to switch failures or driver failures occurring in the CMLI is considered for this comparative analysis. Elapsed time of recovery, Mean Square Error (MSE) and the computational budgets of ANN are the performance parameters considered in this comparative analysis. Optimization is involved in the process of updating the weight and the bias values in the ANN network.  Matlab based simulation is carried out and the results are obtained and tabulated for the performance evaluation. It was observed that Modified Genetic Algorithm performed better than the Genetic Algorithm while optimizing the ANN training.