Distributed model predictive control for district heating networks considering building and network flexibility

The grid-responsive operation of large-scale heat pumps in District Heating Networks (DHNs) requires advanced control strategies to coordinate the DHNs’ inherent thermal storage with power grid demands. The present work introduces a Distributed Model Predictive Control (DMPC) that leverages both building-level and network-level flexibility to adjust heat pump operation in response to dynamic pricing. The DMPC is based on the Alternating Direction Method of Multipliers algorithm to avoid privacy-critical information flows and improve scalability compared to standard centralized MPCs (CMPCs). To achieve this, each building solves a local control problem and exchanges only non-sensitive information with the DHN operator. The DMPC is comprehensively compared against a CMPC on a literature-based benchmark DHN across nine one-week simulation scenarios varying in price and weather conditions. The DMPC successfully achieves control performance comparable to the CMPC with excess costs below 1.5% while maintaining thermal comfort and avoiding privacy-critical information flow. Additionally, the computation time is reduced by 15.5% to 74.3% compared to the CMPC, highlighting the improved scalability. The computation time savings substantially depend on the number of iterations necessary for the DMPC to converge, emphasizing the importance of avoiding excess iterations. These results highlight the potential of distributed control to facilitate scalable, privacy-preserving, and efficient demand response in future DHNs.