Neural Network Based Control of DSTATCOM with Rating Reduction for Three-Phase Four-Wire System

Abstract

Electrical distribution system suffers from problems like reactive power burden, unbalance loading and voltage regulation. Reactive power burden increases line losses and unbalance loading results in excessive neutral current. The use of distribution static synchronous compensator (DSTATCOM) is well established to cater these problems. Two core issues with the use of DSTATCOM are its control and rating. The heart of the control of DSTATCOM is the derivation of reference currents, which decide the switching of voltage source inverter (VSI) used in DSTATCOM. This paper deals with neural network based reference current extractor which operates on LMS algorithm. The neural network is used to extract the real positive sequence current component of load current. To compensate the neutral current, a four leg topology of VSI is used with the dc bus capacitor. The dc voltage across this capacitor is maintained constant using a PI controller. Similarly, in ac voltage regulation mode another PI controller is used to maintain the constant voltage at PCC with variation of load. To reduce the rating of DSTATCOM, a capacitor bank of rating of half the VAR required is connected at the load end. The simulations are carried out in MATLAB environment for unity power factor operation and voltage regulation through leading power factor operation. Simulation results verify the effectiveness of the DSTATCOM system to meet the requirements.