Modified Three-Layered Artificial Neural Network-Based Improved Control of Multilevel Inverters for Active Filtering

Abstract

Multilevel inverter (MLI) especially cascaded H-bridge-type MLI provides transformer-less solution in case of medium-voltage and high-power distribution scenario for mitigating current-related power quality (PQ) problems. Performance of MLI-based shunt filtering unit is dependent on control and extraction technique used. Therefore, a lot of control algorithms are already proposed for conventional two-level inverter-based shunt active power filter (SAPF). However, most control algorithms and their capabilities are yet to be explored for MLI-based SAPF. In view of this concern, this paper aims at presenting an advanced control technique for cascaded H-bridge multilevel inverter-based shunt active power filter. This control scheme composed of three parts, i.e., sensing of source voltage, load and source current, DC-link voltage regulation, and reference current generation technique. DC-link voltage regulation and its characteristics are important in terms of settling time and peak overshoot during load changing condition. Normally, PI compensators are used for DC-link voltage regulation during transient and steady-state conditions. However, soft computing techniques give better response in case of DC voltage regulation. Therefore, modified three-layered artificial neural network (M3L-ANN) is used in this paper for regulating DC-link voltages. The proposed control technique is simulated in MATLAB/Simulink with three-phase and five-level CHB-MLI-based SAPF. Extensive simulation analysis is presented and verified. It shows better performance in terms of settling time and peak overshoot during load changing condition. Simultaneously, source current also becomes well below 5% as per IEEE-519 standard.