Lyapunov Based Neural Network Estimator Designed for Grid-Tied Nine-Level Packed E-Cell Inverter
Mohammad Babaie, Mohammad Sharifzadeh, Majid Mehrasa, Kamal Al‐Haddad
Abstract
This paper presents a robust controller based on Lyapunov Control Theory (LCT) and Artificial Neural Network (ANN) for the grid-tied single-phase nine-level Packed E-Cell (PEC9) inverter. The proposed robust controller is designed using nonlinear model of the grid-tied PEC9 in which a proper negative definite function with a Positive Weighting Factor (PWF) obtained by LCT guarantees the designed controller to provide globally asymptotically stability for the system in unsteady conditions. An online ANN Estimator (ANNE) is also designed to update the PWF which has the most effect on controller performance. ANNE is also trained through an online optimization loop based on Artificial Bee Colony (ABC) algorithm that reduces the training time dramatically and enhances precision of the trained estimator. Some experimental and simulation tests are accomplished by dSpace-1104 hardware and MATLAB/Simulink software to confirm the high accuracy and fast transient response of the proposed LCT-ANNE in the current reference tracking and obtaining low injected current THD for both steady-state and variable conditions of the grid-tied PEC9.