Switching angle optimization and fault analysis of a multistring-multilevel inverter for renewable-energy-source applications

In this paper, a multistring-multilevel inverter (M-MLI) for renewable-energy-source applications has been proposed with reduced switch count and harmonics along with single-switch fault analysis for various levels. It requires only ‘m+1’ power switches for ‘m’ voltage levels. The proposed work achieves the fine-tuning of switching angles using a metaheuristic technique, i.e. the teaching–learning-based optimization algorithm (TLBOA), to mitigate the total harmonic distortion (THD) of the M-MLI. Furthermore, the proposed TLBOA has been compared with conventional modulation techniques such as equal phase (EP), half-equal phase (HEP), near-level control (NLC) and Newton–Raphson (NR) to verify the effectiveness of TLBOA for various voltage levels in terms of % voltage-THD (%V-THD), computational time and methodology. By fine-tuning the switching angles, the %V-THD is improved significantly when compared with EP, HEP, NLC and NR modulation techniques. For an 11-level single-phase M-MLI, the %V-THD using TLBOA at 0.91 modulation index (MI) is 5.051%. The lower-order harmonics, i.e. 5, 7, 11 and 13, are eliminated to improve the power quality. Furthermore, MLIs are often prone to failure, resulting in waveform distortion. The extreme reduction in power quality impacts the load and significant damage is likely. The location of the open-circuit fault to be identified becomes more tedious under the faulty conditions with increased switch counts and voltage levels since the mathematical modelling fails to address the scenario in less computational time. Hence, the machine-learning approach, i.e. support vector machine (SVM) with Bayesian optimization, has been discussed to locate the faulty switch. Finally, the proposed M-MLI configuration has been modelled, simulated and validated using MATLAB® and Simulink®. The results of the M-MLI configuration have been verified for 7, 9 and 11 levels using TLBOA along with fault analysis using the SVM approach.