Leveraging Grey Box Models for Enhanced Energy Flexibility in Centralized and Decentralized Single-Zone Multi-Node Systems

Ozan Baris Mulayim, Mario Berges

Abstract: The effectiveness of residential energy flexibility (EF) largely depends on maintaining occupant comfort to prevent the rendering of events ineffective through overrides. Model Predictive Control (MPC) has demonstrated significant achievements in avoiding discomfort while providing EF, notably through the deployment of Grey Box Models (GBMs) for scalable thermal behavior modeling. With the advent of smart thermostats and remote sensors, residential buildings now have the potential to benefit from sophisticated thermal behavior modeling techniques previously utilized for commercial buildings. Yet, neither the efficacy of GBMs in these settings nor the impact of temperature variances from various sensors on EF has been thoroughly investigated. Therefore, this study explores centralized and decentralized GBMs for Single-Zone Multi-Node Systems (SZMNSs), aiming to accurately represent node-level thermal dynamics for enhancing EF predictions. Utilizing data from 1,000 smart thermostats, our research focuses on three primary objectives: (1) the derivation of physics-based equations for SZMNSs equipped with sensor networks and smart thermostats, (2) a comparison of the predictive performance of four different modeling paradigms, and (3) the exploration of node-level EF potential within SZMNSs using our developed models. The results affirm the applicability of multi-zone modeling techniques to SZMNSs with errors less than 1°F and highlight that the EF predictions can be 7.6% higher or 29.1% lower based on which node is used for control. The effective parameterization and deployment of these models demonstrate their potential for facilitating demand-responsive, energy-efficient solutions in residential contexts, with negligible impact on occupants’ thermal comfort.