Optimization of a Solar-Powered DC Microgrid for a Rural Community Using AI-Based Load Forecasting

Abstract

This paper presents the design, simulation, and AI-based optimization of a Solar-Powered DC Microgrid for rural electrification. Modelled in MATLAB/Simulink, the system integrates a Solar PV array (5 kW), Wind Turbine (10 kW), PEM Fuel Cell (28 kW), Battery Storage System (50 Ah / 700 V), bidirectional DC-DC converters, AC/DC rectifiers, and a three-phase grid interface-all interconnected through a common 700 V DC bus. Maximum Power Point Tracking (MPPT) with a Pertur-band-Observe (P&O) algorithm optimizes solar extraction, while the novel Adaptive Learning Algorithm based Tuned Kalman Adaptive Network (ALA-TKAN) controller provides real-time adaptive optimization of energy dispatch. PWM-based duty-cycle control regulates converter/inverter operation. Simulation results over a 5-second window demonstrate improvements of 30-35% in efficiency (90%), power quality (88%), stability (85%), and reliability (92%) over conventional single-source systems, confirming suitability for rural and off-grid applications.