Neural network and polynomial model to improve the coefficient of performance prediction for solar intermittent refrigeration system

This study presents a novel hybrid methodology to estimate the coefficient of performance in an absorption intermittent cooling system; the system is for ice production and operates with an ammonia/lithium nitrate mixture. The hybrid model integrates a polynomial fitting method and an artificial neural network model to improve the network performance and the estimation of the COPs. The improvement uses fewer hidden neurons without sacrificing accuracy in the prediction. The proposed hybrid model has two neurons in the input and two in the hidden layers and shows better results than those obtained through polynomial fitting or artificial neural networks separately. The developed model presents an excellent agreement between experimental and simulated values of the coefficient of performance with a determination coefficient R2>0.9978.