ANN-Based Thermal Load Prediction Approach for Advanced Controls in Building Energy Systems

Abstract

The Artificial Neural Network (ANN) technology has been used in various areas. In the building industry, however, ANN is relatively less utilized due to its complexity and uncertain benefits of its application along with the costs associated with its development. This paper introduces ANN regarding its applicability and potential benefits in building operations, especially for energy savings. Thermal loads calculations are most widely used for the operation of building energy systems. An ANN model was developed to predict a large office building's cooling loads. The EnergyPlus simulation program was used to generate thermal loads data and the Python program to develop an ANN model. The initial ANN model predicted a case study building's cooling loads within the CVRMSE value of 7.3% initially, and later 6.8% after optimization, which is within the tolerance range of 3Q% recommended by the ASHRAE Guideline 14. This study showed the potential benefit of energy savings that can be achieved by utilizing the ANN model for accurately predicting the cooling loads.