Integrated neural network-based MPPT and ant colony optimization-tuned PI bidirectional charger-controller for PV-powered motor-pump system

Abstract

This study presents the design and implementation of an efficient off-grid photovoltaic (PV)-powered motor-pump system utilizing a two-stage power converter. The system integrates a neural network-based maximum power point tracking (MPPT-NN) algorithm with a proportional integral (PI) controller and an additional bidirectional PI charger. Controller gains are optimized using ant colony optimization (ACO) to achieve optimal performance. The proposed MPPT-NN-PI/ACO controller enhances control responses and improves energy utilization efficiency by 17% compared to traditional PI controller. Performance comparisons of MPPT techniques demonstrates that the proposed controller outperforms several existing methods, including commercial on-off controllers, the modified Perturb & Observe algorithm, and neural network-based controllers, by approximately 4%–20%. It shows a slightly different performance of about 1%–6% compared to advanced adaptive controllers, including fuzzy logic and neuro-fuzzy controllers. For bidirectional charger performance, the DC bus voltage connecting the boost converter and bidirectional converter remains stable with small ripples and is well-aligned with the reference voltage, ensuring uninterrupted operation under varying weather conditions. The bidirectional charge management effectively maintains battery state-of-charge (SOC), showing a decline during periods of insufficient PV energy and achieving full charging during periods of excess PV energy. System performance is validated through both simulation and laboratory-scale prototyping, ensuring robust operation.