The wide range of methodologies employed in Energy Management Systems (EMS), spanning rule-based strategies, heuristic approaches, optimization algorithms, model-based frameworks such as Model Predictive Control (MPC), and data-driven techniques including machine learning and reinforcement learning, necessitate a thorough evaluation of their performance and adaptability. Yet, despite the diversity of methods, there is a lack of systematic benchmarking across paradigms, which leaves stakeholders without clear guidance on selecting suitable control strategies. This paper conducts a comprehensive benchmarking of these methodologies to identify the most effective approaches, highlighting their strengths, limitations, and potential for driving future innovations. By focusing on critical performance metrics such as energy efficiency, bill reduction, and scalability, this work aims to provide valuable guidance to stakeholders in selecting and implementing optimal energy management solutions for the smart buildings and energy grids of the future. Our experimental results show that reinforcement learning methods, particularly PPO and DQN, consistently outperform rule-based and MPC approaches in terms of both electricity bill reduction and PV self-consumption, demonstrating their strong potential for future iEMS deployment.