Artificial Intelligence (AI) has emerged as a transformative paradigm in electrical power systems, enabling
real-time optimization, predictive control, and intelligent energy distribution. This paper proposes a unified AI-driven
framework for power management integrating machine learning, deep reinforcement learning, and edge intelligence
across smart grids, microgrids, and distributed energy resources (DERs). The proposed architecture enhances
efficiency, resilience, and sustainability by enabling adaptive load balancing, predictive maintenance, and dynamic
demand-response control. A comparative analysis demonstrates that AI-based systems can achieve significant
improvements in forecasting accuracy, operational cost reduction, and system stability. Recent studies show load
forecasting errors can be reduced below 2% using advanced AI models. The paper further explores challenges such
as computational overhead, data integrity, and cybersecurity. Finally, future research directions involving federated
learning, digital twins, and AI-integrated energy markets are discussed.