Multi-objective online optimization of building energy systems for improved control smoothness and efficiency

Conventional optimization algorithms face challenges in their practical applications to online optimization due to a lack of control smoothness, particularly for building energy systems. Therefore, this paper proposes a multi-objective framework for online optimal control of building energy systems to achieve both smooth and energy-efficient control. The framework treats the distance between successive control actions as a co-equal optimization objective alongside energy efficiency, generating a Pareto front to explicitly map the trade-off between control smoothness and cost. A user-adjustable tolerance level is then employed to select a solution from the Pareto front for online control. The proposed framework is validated on the optimal chiller loading problem in a four-week data experiment. Compared to the best baseline algorithm in the experiment, differential evolution (DE), the framework achieves significant enhancement in control smoothness, as evidenced by an 18.9 % reduction in the total chiller switching number without sacrificing energy efficiency.