Multi-Source Building Energy Load Forecasting using a Transformer-Mamba Framework for Sustainable Infrastructure
Abstract
Accurate forecasting of multi-source building energy loads is an important requirement in modern civil engineering for efficient building operation, infrastructure resilience, and sustainable resource utilization. Conventional forecasting approaches often experience limitations in representing nonlinear demand variations, long-term temporal dependencies, and interactions among multiple utility systems such as electricity, chilled water, steam, and hot water. Several existing methods also encounter computational inefficiency when applied to large-scale hourly building datasets. To address these challenges, this study proposes a hybrid Transformer Encoder–Mamba framework for multi-source building energy load forecasting in sustainable building operation. The Transformer Encoder is utilized to identify global relationships among climatic, temporal, and operational variables, while the Mamba Selective State Space Model (SSM) captures extended sequential behavior with improved computational efficiency. This integration enables effective modeling of complex feature interactions alongside long-horizon temporal dependencies in large-scale building datasets. The proposed framework was validated using the ASHRAE Great Energy Predictor III dataset containing three years of hourly meter readings, weather records, and building metadata from more than one thousand buildings across multiple international sites. Experimental results achieved a Root Mean Square Error (RMSE) of 0.071, Mean Absolute Error (MAE) of 0.054, Mean Absolute Percentage Error (MAPE) of 4.38%, and Coefficient of Determination (R2) of 98.9%. Furthermore, comparative evaluation with recent state-of-the-art forecasting models and additional validation experiments confirmed the effectiveness, robustness, and generalization capability of the proposed framework. The findings demonstrate its suitability for smart building design, energy-efficient facility management, and sustainable urban infrastructure planning.