Modelling and experimental validation of DSTATCOM using a deep belief learning network with an anti-wind-up regulator

Abstract

This article proposes the shunt compensation capability improvement using a deep belief learning network approach (DBLN) with anti-wind-up regulator-supported distributed static compensator (DSTATCOM). Six subnets make up this proposed DBLN controller. Three subnets for each active and reactive mass part are employed to isolate the basic component of the output current. Numerous issues such as past and normalising weight and learning rates are engaged in the DBLN-based weight-updating formula to have a superior dynamic presence, reduce the computational load and achievesfaster estimation, etc. This proposed DBLN is suggested for both proportional-integral (PI) and anti-wind-up regulator to showcase the better DC link voltage which further leads to providing better PQ improvement. This method offers excellent dynamics and resilience to outside disturbances. The suggested study is examined by simulation and experimental development using MATLAB/Simulink by a real-time interface based on a dSPACE 1104 for healthier potential regulation, potential balancing, input current harmonic distortion and PF correction under different load scenarios.

KEYWORDS:

• DSTATCOM
• DBLN
• anti-wind-up regulator
• PQ

Disclosure statement

No potential conflict of interest was reported by the author(s).