Experimental analysis of artificial intelligence-based model predictive control for thermal energy storage under different cooling load conditions

Abstract

As the global community seeks to reduce fossil fuel dependence, improving energy efficiency and developing an optimal operational strategy are high priorities in the building sector. Integrating thermal energy storage (TES) into building energy systems is desirable for improving energy flexibility and cost efficiency. However, an optimal control strategy for TES system operation is required to maximize its advantages of balancing energy supply and demand. This study proposed a model predictive control (MPC) scheme using artificial intelligence (AI) to control the TES system and validated it by the experimental analysis. To construct a reliable and computationally manageable MPC controller, an artificial neural network was utilized as a prediction model, and a metaheuristic algorithm of εDE-RJ was employed as an optimization solver. The AI-based MPC for cooling operation was examined using a reduced-scale experimental system consisting of a chiller, sensible TES tank, heat exchangers, and variable-speed pumps. AI-based MPC optimized the charging and discharging rates of the TES system to minimize the operating cost. Three cooling load profiles were tested for both MPC and rule-based control methods. The results showed that AI-based MPC was highly feasible by flexibly managing the different load levels and reducing the operating cost by 9.1–14.6%.