Supervisory optimization strategy for energy minimization in multi-evaporator air-conditioning systems

초록

Multi-evaporator air-conditioning (MEAC) systems exhibit strong multivariable coupling and wide operating envelopes, complicating supervisory energy optimization under real-time constraints. This paper proposes a computation-light condenser-fan optimizer that exploits the trade-off between reduced compressor power at lower condensing pressure and increased fan power at higher fan speed. A control-oriented predictive model is built using a moving boundary method, enabling rapid prediction of total outdoor-unit power across changing conditions. Golden section search is applied, where the method is derivative-free, robust, and implementationfriendly. The resulting supervisor converges to a near-optimal fan command with a small, deterministic number of model evaluations, making it suitable for online execution. The approach is validated against experimentally measured power trends of MEAC equipment and benchmarked against dense fan-speed sweeps. Results show that the optimized solution closely matches the experimentally observed minimum and reduces unnecessary fan operation under mild ambient conditions without sacrificing stable high-pressure control. The proposed framework provides an interpretable and deployable pathway to real-time energy minimization in practical embedded controllers. Compared with the conventional fixed high-pressure setpoint policy, the proposed supervisor achieves up to 9.6% measured power savings on the physical 29 kW MEAC unit, with computation time below 300 s, supporting real-time supervisory deployment.

키워드