System-level multi-objective optimization of a magnetic air conditioner through coupling of artificial neural networks and genetic algorithms

Abstract

This work advances an integrated approach to design the components of a magnetocaloric air conditioner capable of producing a cooling capacity of 9000 BTU h−1 (2637 W) for indoor and outdoor ambient temperatures of 22 °C and 35 °C, respectively. Through a combination of data-driven and deep learning-based approaches, standalone models were developed for the active magnetic regenerator and magnetic circuit. The cooling system analysis was complemented by existing models for the tube-fin heat exchangers and the hydraulic management system. A multi-objective method based on genetic algorithms was proposed to optimize the system for minimal power consumption and the total assembly cost. The final design exhibited a COP of 2.57 and a second-law efficiency of 11.4%. A power consumption breakdown revealed that the most significant contributions are the magnetization of the solid refrigerant and fluid pumping. As for the capital cost and system mass, the most critical contribution is associated with the magnetic circuit’s rotor due to the volume of soft and hard magnetic materials.