Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : July 08, 2021 - July 09, 2021
Air conditioning is a well-established technology, which is broadly implemented all over the world in different system sizes and working conditions, but no conclusive performance prediction approach is presently available. An effective method for the performance prediction of operative air conditioners is developed by taking advantage of the benefits of machine learning techniques while overcoming the main limitations in generality of the results obtained in previous related literature. The ability of Artificial Neural Network (ANN) in reconstructing complex interrelations between physical parameters of interest is investigated with reference to different categories of input quantities for the prediction of the actual performance of the system. As vapour compression air conditioners share the same fundamental working principle, it is demonstrated that generalizable and accurate predictions are possible by training the ANN with refrigerant temperatures which are representative of the operating cycle. Accordingly, the proposed method represents a cost-effective, non-intrusive, and accurate option for effective energy management procedures.
