Regression tree ensemble learning-based prediction of the heating and cooling loads of residential buildings

Abstract

Building energy consumption is heavily dependent on its heating load (HL) and cooling load (CL). Therefore, an efficient building demand forecast is critical for ensuring energy savings and improving the operating efficacy of the heating, ventilation, and air conditioning (HVAC) system. Modern and specialized energy-efficient building modeling technologies may offer a fair estimate of the influence of different construction methods. However, deploying these tools could be time-consuming and complex for the user. Thus, in this article, an ensemble model based on decision trees and the least square-boosting (LS-boosting) algorithm known as the regression tree ensemble (RTE) is proposed for the accurate prediction of HL and CL. The hyper parameters of the RTE are optimized by shuffled frog leaping optimization (SFLA), which leads to SRTE. Stepwise regression (STR) and Gaussian process regression (GPR) based on different kernel functions are also designed for comparison purposes. Results demonstrate that the value of root mean squared error is reduced by 37%-68% and 30%-41% for HL and CL of residential buildings, respectively, by the proposed SRTE in comparison to other models. Furthermore, the findings from the real dataset support the proposed model's effectiveness in predicting HVAC energy usage. It can be concluded that the proposed SRTE is more effective and accurate than other methods for predicting the energy consumption of HVAC systems.