Hybrid DSTATCOM Design Using Deep Belief Networks for Enhanced Power Quality Assessment

Abstract This article proposes the power quality (PQ) assessment using a deep belief-learning network (DBLN) approach-based inductor and capacitor (LC) supported distributed static compensator (DSTATCOM). This suggested DBLN controller is constituted by considering six sub networks for direct and quadrature components of three phases. Several factors like previous weight, step size, harmonic component and learning rate are associated in the DBLN learning mechanism to possess better dynamic performance. This proposed DBLN is suggested for both DSTATCOM and LC coupled DSTATCOM to showcase the proper DC link voltage regulation, which furthermore leads to providing better PQ improvement. To build a high-accuracy evaluation model LC coupling is analysed and designed by means of mathematical analysis and incorporated in the system. The proposed study is investigated by simulation and practical implementation using MATLAB/Simulink and hardware setups to improve power factor (p.f.) correction, source current harmonic reduction, voltage balancing and voltage control under various loading scenarios as per IEEE-519-2017 and EN-50160.