A Hybrid Fuzzy Predictive DVR Model for Voltage Estimation Using Intelligent Learning

A self-constructing intelligence-based learning law is proposed for the prediction of power quality mitigation through a dynamic voltage restorer (DVR). To tackle voltage-based power quality concerns in the distribution network, a fuzzy neural-based hybridization control approach using an evolving Takagi–Sugeno–Kang (e-TSK) and radial basis function (RBFN) is developed. The estimated value of fundamental weight components is achieved by the eTSK predictive model. The evaluated weight values are used to update the parameters for the generation of the reference load voltage components. The RBFN is used to achieve the desired task of dc and ac link voltage regulation. The self-tuned control algorithm is developed based on an intelligence technique that avoids the trapping of local optima, and early convergence, and reduces the voltage oscillation to obtain an optimal solution. A meta-heuristic optimization approach of grey wolf optimization (GWO) and grey wolf-cuckoo search (GWOCS) is selected to optimize the parameters of eTKS and RBFN to find the best-fitted predictive model under the dynamic state. The hybridization of the model with GWO and GWOCS is implemented for improving the training and generalization capability. The performance of VSC-based DVR is validated using the d-SPACE Micro lab box under the various power quality scenario.