A Self-Tuning ANFIS DC Link and ANN-LM Controller Based DVR for Power Quality Enhancement

Abstract

An artificial intelligence integrated control is proposed for a three-phase dynamic voltage restorer (DVR). The proposed Levenberg-Marquardt back-propagation (LMBP) algorithm is developed by employing intelligent computational system under supervised learning. The optimized artificial neural network (ANN) model is used for the fundamental computation of load voltage components through the training process. The common training problem of ANN models are slow learning of system and get trapped in local optimum. The proposed LMBP hybridized learning system reduces the error rate and taking this advantage it overcomes the aforesaid issue. It is integrated with the adaptive neuro-fuzzy inference system (ANFIS) controller for regulating the DC and AC link voltage error. In the proposed design of ANN-AN-FIS based DVR, a hybrid learning algorithm and Gaussian membership functions are applied to extract the best forecasted ANFIS models. The trained model accuracy is evaluated based on statistical indices. The obtained values during the training state for DC link voltage error regulation are mean square error $(\pmb{MSE}=0.00054585$ ), standard deviation ( $\pmb{SD}=0.023364$ ), and regression ( $R=1$ ) are found effective for the approximation of ANN-LMBP model. The simulation results obtained from ANN-LMBP, and ANFIS models were tested on Micro-Lab Box experimentally which shows the improved power quality response at various operating conditions.