Efficient Energy Management in Grid-Connected Microgrids with Battery Storage

Abstract

The global rise in electricity demand driven by electrification and industrialization has led to frequent power system failures, with grid overloading during peak demand periods posing a major risk of collapse. To address this challenge, this study develops an optimal energy scheduling strategy for a grid-connected micro grid using an Artificial Neural Network (ANN). The ANN model was trained on 70% of a dataset of 1024 load variation scenarios, with 30% reserved for testing and validation, and incorporated total system load and battery state of charge as inputs to determine operational decisions such as switching between solar and wind power and managing battery charging/discharging. The proposed micro grid integrates solar and wind energy with the IEEE 14-bus test system, supported by battery storage to maintain supply-demand balance. Simulation results show that the photovoltaic system generated 32.9 MW (35% of the critical 94 MW load), while the wind system produced 61.1 MW from 82 turbines (65% of the load). The system was tested under normal, underload, and overload conditions, and ANN performance was further enhanced by Particle Swarm Optimization (PSO), yielding a 24.21% improvement compared to standard ANN training. The entire framework was modeled in MATLAB/Simulink, and validation under three operational scenarios confirmed the reliability and effectiveness of the proposed scheduling approach in improving grid efficiency and supporting renewable energy integration.