Performance Evaluation of Photovoltaic Based Distributed Generation against Neural Network based Feedback Control

Abstract

The Independent Distributed Generator Controller (IDGC) design is emphasized here for Photovoltaic (PV) based applications. The PV based Distributed Generator (DG) are required to have Maximum Power Point Tracking (MPPT) control with provision of closed-loop feedback path, especially for grid interactive operations. The Neural Network based MPPT control algorithms are quite accurate to estimate converter's input Control References (CRs: voltage/ power at MPP). The Feedback controls are well-established with linear Proportional-Integral (PI) scheme or non-linear, complex techniques. This conventional two-step converter controller/ IDGC approach is improved in this paper, with direct estimation of converter output CRs (e.g. duty cycles, PWM index). A fast learning (Moore-Penrose pseudo-inverse) based Extreme Learning Machine (ELM) is considered with new, online-sequential, ridge re method towards robust estimation of CRs. The accurate CRs for considered PV-DG is evidenced towards improved stability, under PV side as well as grid uncertainties. The validation of proposed IDGC operation is performed in MATLAB environment.