Thermal allocation strategy for building complexes based on distributed model predictive control: Application to a solar district heating system

Solar district heating (SDH) systems are a pivotal technology for the low-carbon transition of space heating, but centralized model predictive control (CMPC) suffers from high computational burden and poor scalability when many buildings are coupled through large thermal networks. This study proposes a two-layer distributed model predictive control (DMPC) framework with ADMM, heuristic acceleration, and a Python–TRNSYS co-simulation platform for SDH systems with multiple building complexes. A representative 168 h winter scenario and scalable test cases are used to benchmark DMPC against CMPC. The results show that DMPC reduces the indoor temperature standard deviation by 12–18 % and decreases the maximum inter-complex temperature difference under disturbances from 1.92 to 0.85 °C. The mean absolute error between heating supply and load is reduced by 65.4–84.8 % across the four building complexes. At the system level, DMPC increases collected solar heat by 7.1 % and reduces auxiliary heat supplied by the electric boiler by 13.6 %. Moreover, DMPC reduces total electrical energy consumption of pumps and auxiliaries by 9.0 % relative to CMPC. In a 64-building scenario, the distributed formulation achieves a 335-fold speed-up and reduces memory usage to 3.4 % of that required by CMPC. These findings demonstrate that the proposed DMPC architecture can simultaneously enhance comfort, energy efficiency, and computational scalability for large-scale SDH systems.