• 2025-04-25
• K.A. Hassan, M.K. Agboola, T.O. Ajewole, O. Olukayode
• Volume 1, Issue 1

Abstract

This study presents a comparative analysis of two control strategies; Proportional-Integral (PI) control and a Reinforcement Learning Brainstorm Optimization (RLBSO)-trained Artificial Neural Network (ANN) controller—for improving power quality and reducing losses in hybrid-source microgrids under voltage disturbances. Simulated conditions included voltage sag, swell, and combined disturbances over a 1-second period. In the absence of control, active power losses reached 2,000 W and reactive losses peaked at 1,000 Var, with an overall system loss of approximately 20%. PI control reduced active losses to 1,400 W (30% improvement) and reactive losses to 700 Var. In contrast, the RLBSO-trained ANN controller significantly improved performance, reducing active losses to 600 W (70% reduction) and reactive losses to 200 Var (80% reduction), while keeping power variation within ±5%. Additionally, the ANN-RLBSO controller achieved faster stabilization—settling grid power at 15,000 W and load power at 145,000 W—whereas the PI controller showed slower convergence and larger oscillations. These results demonstrate the ANN-RLBSO controller’s robustness, adaptability, and efficiency in maintaining stable power quality under dynamic disturbances. The method offers a promising solution for intelligent control in hybrid microgrids. Future work will explore real-time deployment and expanded optimization strategies.

Download PDF