Developing machine-learning meta-models for high-rise residential district cooling in hot and humid climate

Cooling accounts for a significant amount of energy consumption in hot and humid climates, and district cooling is an energy-efficient solution. During its planning stage, an accurate and fast prediction of district cooling loads is required to assist decision-making. This study investigated the development of machine-learning-based meta-models to predict the cooling load of high-rise residential buildings at a district scale. Four machine-learning models have been evaluated, including Multiple Linear Regression, Support Vector Regression, Artificial Neural Networks (ANN), and eXtreme Gradient Boosting. The meta-model development starts with the sensitivity analysis to determine the critical parameters as independent variables. The results show that the ANN model with 30 neurons of the hidden layer, trained by eight epochs with 20% sample data, manifested superior performance when predicting monthly energy use intensity for the testing cases. A minimum of eight independent variables for the meta-model was also shown to be acceptable.