Power-constrained VRF system optimization using symbolic regression for multiple zones environment

The rapid growth in global energy consumption highlights the urgency of doubling energy efficiency improvements by 2030. Heating, ventilation, and air-conditioning (HVAC) systems, which account for nearly half of building energy use, represent a critical target for optimization. Conventional HVAC control strategies, however, often suffer from inefficient power allocation, high peak demand, and compromised thermal comfort, especially under dynamic occupancy and environmental conditions. Existing multi-zone control methods often overlook peak power constraints and are not designed to optimize energy use under variable occupancy conditions, resulting in suboptimal energy performance. This study proposes a symbolic regression-based model predictive control (MPC) framework to address these challenges. The framework optimizes energy consumption and thermal comfort for multi-zone variable refrigerant flow (VRF) systems while addressing peak power constraints to reduce energy costs and improve thermal comfort. The method is evaluated under three operating priorities, ω = 0.1, 0.5, and 0.9, across varying power constraints. Simulation results demonstrate that the proposed method consistently outperforms a decentralized MPC state-of-the-art ( SOTA ) baseline, achieving up to 16 % energy savings under a 30 % power constraint, with average temperature deviations (ATD) remaining within comfortable bounds ( < 2 ∘ C ). Even under tight energy constraints, the framework maintains stable control performance, outperforming existing methods that fail to adequately manage peak loads. Compared to rule-based and model-based MPC approaches, the proposed method is more flexible and robust, as it does not require detailed system identification or extensive training data. These results highlight the method’s potential as a scalable and energy-efficient solution for contributing to global energy efficiency goals.