Controlling the Direct Connected Parallel Three-Phase Voltage Source Inverters by Using Neural networks

Abstract

This paper demonstrates that neural networks can be used effectively for the controlling the direct connection of parallel three-¬phase voltage source inverters..A unique feature of the parallel three-phase inverters is a zero--sequence circulating current. This work proposes a new zero-sequence control for parallel three- phase voltage source inverter using a neural networks to assist the Proportional-Integral controller against the variation on the two inverters to maintain the value of a zero-sequence current within an acceptable value. The controller of the zero-sequence current can be implemented within an individual inverter and it is independent of the other control loops of the inverter, therefore, it greatly facilitates design and expansion of parallel system.Simulation and results can be obtained by using mathematical software such as MATLAB version seven to show the performance of the controllers and the closed loop system.Two-layers feedforward neural networks FNN's containing a Levenberg-Marquardt training algorithm are used.