Modeling baseline electrical energy use of chiller system by artificial neural network

Abstract

Artificial Neural Network (ANN) is a nonlinear signal processing devices and works based on human brain principles. It is one of the most widely used techniques for forecasting and predicting. In this work, ANN is used to model baseline electrical energy use for a chiller system. In this work, three inputs that are affecting energy use of the chiller system are selected i.e. 1) refrigerant tonnage, 2) inlet temperature and 3) outlet temperature. Meanwhile the output is electrical energy use. The ANN model is simulated with 81 different structures for both single and two hidden layers. The best ANN structure is selected in order to develop the baseline. Three different performance functions are used to check the model accuracy which are coefficient of correlation (R), mean square error (MSE) and mean absolute percentage error (MAPE). Results show that ANN gives higher prediction performance with the value of R is above 93%, small error value of MSE and MAPE for the selected ANN structure.