Advancing energy optimization in building systems through multi-scale sensitivity, adaptive modeling, and real-time management

• Evaluates building energy enhancement using Bayesian Adaptive Spline Surfaces (BASS) across multiple time scales. • Confirmed a systematic procedure for global variance-based sensitivity analysis on annual, monthly, and daily. • Utilizes BASS models for fast and reliable sensitivity results, essential for multi-time-scale energy assessments. • Incorporates principal component analysis to manage high correlation among multi-output energy uses at finer scales. • Offers a scalable framework applicable to other energy systems, highlighting the versatility in building energy analysis. This study provides a comprehensive assessment of building energy optimization through a systematic methodology utilizing Bayesian Adaptive Spline Surfaces (BASS) across multiple time scales. The complexity of sensitivity analysis increases when addressing issues embedded in multiple time scales. To address this, the proposed approach extends global variance-based sensitivity analysis to annual, monthly, and daily scales. Case studies demonstrate how the methodology leverages BASS models to achieve rapid and precise sensitivity outcomes essential for multi-time-scale energy evaluations. BASS models offer significant advantages, including high predictive accuracy and efficient computation of sensitivity metrics without relying on Monte Carlo integrals. Additionally, the integration of principal component analysis helps manage the high correlation among multiple energy outputs at finer scales, leading to substantial computational cost savings. The findings reveal that this sensitivity analysis approach provides new insights into the energy characteristics influenced by varying weather conditions across different time scales. The results establish a scalable framework applicable to diverse energy systems and time scales, showcasing the adaptability and effectiveness of this method in advancing the field of building energy analysis.