Time allocation of a three-bed adsorption chiller using an artificial neural network. (December 2021)

• Record Type: Journal Article
• Authors: Lee, Woo Su; Park, Moon Yong; Duong, Xuan Quang; Koushaeian, Mahdi; Shah, Nehad Ali; Chung, Jae Dong
• Is Part Of: Case studies in thermal engineering. Volume 28(2021)
• Publication Date: 2021-12
• Subjects: Adsorption chiller -- Time allocation -- Artificial neural network; Heat engineering -- Case studies -- Periodicals; 621.40205
• DOI: 10.1016/j.csite.2021.101553
• Languages: English
• ISSNs: 2214-157X

Abstract

Abstract: 3-bed adsorption cooling systems have the advantage of maintaining a uniform chilled water temperature. Because the number of parameters in a 3-bed system is very large, using artificial neural networks (ANN) is often suggested as an alternative to conducting experiments in many recent studies. We systematically determined the optimal time allocation for a 3-bed, 2-evaporation adsorption cooling system, using ANN with five variables, i.e., adsorption/desorption time ratio ( f ad ), high/low evaporator time ratio ( f p ), cycle time (τ), and the time lag between each adsorption bed ( δ 2, δ 3 ). Each case was methodically modeled using a process that our research team had previously developed and verified. When the coefficient of performance (COP) and specific cooling power (SCP) estimated by ANN were compared with the actual results, errors were within ±4%. Finally, the best strategy for each performance indicator, i.e., COP, SCP, and standard deviation of chilled-out temperature, was proposed for the 5 operating parameters of f ad, f p, τ, δ 2, and δ3 . The cycle time was found to have 42.9% relative importance for the COP, and 33.1% relative importance for the SCP, but the most influential factor to the SCP was the high/low pressure evaporator time ratio with a relative importance of 33.3%. Highlights: A 3-bed, 2-evaporation adsorption cooling system was analyzed. The time allocation was systematically determined using artificial neural networks (ANN). Using artificial neural networks with 5 variables were examined for the system performance. The impact of cycle time on COP was determined to have a relative significance of 42.9%. The impact of cycle time on COP was determined to have a relative significance of 33.1%.