AI-BASED PREDICTIVE ENERGY MANAGEMENT OF INTEGRATED MEP SYSTEMS IN GRID-INTERACTIVE COMMERCIAL BUILDINGS

Abstract

This study examined an AI-based predictive energy-management framework for integrated mechanical, electrical, and plumbing (MEP) systems in Pakistani commercial buildings. A quantitative, predictive, quasi-experimental design was employed to evaluate short-term energy-demand forecasting, coordinated MEP control, peak-demand management, and grid-interactive energy flexibility. The framework integrated building energy, weather, occupancy, and operational data with machine-learning models and an optimization layer to support predictive rather than reactive energy management. The results indicated that the hybrid AI model achieved strong forecasting performance (R² = .97; MAPE = 3.91%). AI-managed operation was associated with a 12.34% reduction in electricity consumption, a 14.79% reduction in peak demand, and a substantial increase in load-shifting capacity. Regression results further indicated significant relationships between AI-based predictive management, MEP integration, energy efficiency, forecasting accuracy, and energy flexibility. The study contributes to the Cyber-Physical Systems perspective by linking real-time sensing, AI prediction, optimization, and physical MEP control within a closed-loop framework. The findings highlight the potential of AI-enabled predictive management to improve commercial-building energy efficiency and grid responsiveness in Pakistan.