SCALABLE, DATA-DRIVEN, AND HIGH-RESOLUTION BUILDING MODELS FOR CONTROL OPTIMIZATION

Model predictive control of building HVAC systems has the potential to reduce HVAC energy use by 20% to 40% over existing control strategies, with minimal capital cost. Unlike existing HVAC control strategies that rely on a set of pre-programmed rules, Model predictive control continuously evaluates all possible control strategies to choose the one that minimizes current and future energy use while meeting comfort, equipment, and other constraints. While Model Predictive Control has been studied extensively in academic literature, commercial deployment of MPC beyond demonstration projects remains limited. The primary obstacle to widespread commercial deployment is the engineering costs associated with training and maintaining the building models suitable for Model Predictive Control at scale. eCurv proposes development of improved building models and modeling processes for Model Predictive Control that require minimal engineering effort to construct and maintain, that are applicable for multiple building configurations, and that capture fast-acting transient effects. This development in building modeling capability will enable Model Predictive Control to become a commercially viable, by decreasing the engineering costs associated with training and maintenance and will broaden the application of Model Predictive Control in areas such as peak demand management and grid ancillary services. In Phase 1, eCurv aims to evaluate the technical feasibility and merit of candidate models and processes, based on a combination of real building data and simulations. eCurv aims to evaluate models for ease of training and implementation, temporal resolution and ability to capture transient effects, and generalizability and ability to adapt to changing building conditions. eCurv also aims to evaluate the commercial feasibility of candidate models, processes, and associated business/revenue models.