A Hybrid AI and Predictive Control Framework for an IEEE 33-BUS Microgrid Energy Management System

Abstract

The increasing penetration of distributed energy resources, battery energy storage systems, electric vehicles, and flexible residential loads has intensified the need for adaptive and intelligent energy management systems in modern distribution networks. Existing studies have addressed different aspects of this problem through metaheuristic DER planning, multi-objective optimization, forecasting-based EMS, model predictive control, reinforcement learning, and smart inverter voltage regulation. However, most of these contributions treat planning, dispatch, and adaptive control as separate layers, often relying on static operating snapshots, fixed controller weights, simplified power-flow models, or limited validation of voltage, thermal, and storage constraints. This thesis proposes an integrated AI-based EMS framework for the IEEE 33-bus radial distribution network by combining metaheuristic DER placement and sizing, BESS coordination, Volt-VAR support, MPC-based operational control, and RL-based adaptive tuning. The proposed framework connects long-term planning decisions with short-term operational control under nonlinear AC power-flow validation. By embedding voltage limits, line-loading constraints, and battery state-of-charge dynamics into the EMS evaluation, the work aims to provide a physically grounded and technically interpretable approach for improving loss reduction, voltage performance, and operational flexibility in active distribution networks.