Abstract: Aiming at the problems of significant differences in user operation habits and insufficient energy-saving awareness in heat pump heating systems, it proposes a Model Predictive Control (MPC) strategy for heat pumps based on a data-driven hybrid model. By analyzing the operational data of 210 users in northern China, user behaviors were categorized into three types: non-adjusting, less-adjusting, and frequent-adjusting, revealing that 60% of users exhibit operational inertia or a lack of knowledge. On this basis, a hybrid model combining a 4R3C building thermal resistance-capacitance model with a BP neural network for heat output prediction was constructed. Parameters were identified using a genetic algorithm, and automatic optimal adjustment of water temperature was achieved through MPC. Experimental results show that this strategy can achieve energy savings of 6.5% to 15.7% while maintaining indoor comfort, verifying its effectiveness and feasibility in practical applications.

Key words: Heat pump, Model predictive control(MPC), Data-driven, Hybrid model, Energy-saving operation, User behavior analysis