Volume 66

Comparison of Prediction Algorithms for Heating Building Mean Room Temperatures Yongjie Wang, Guanjing Lin, Changhong Zhan

Abstract

Accurately predicting the trend of mean room temperature under known operational states of heating systems is of great significance for achieving low-carbon and energy-efficient regulation of heating systems. Existing studies have not clearly identified the most suitable method for predicting building mean room temperature. Therefore, the effectiveness of various machine learning algorithms, mathematical regression algorithms, and recently emerging bio-inspired optimization algorithms in predicting the mean indoor temperature of buildings was compared and analyzed. Data were collected from an office building, and the input variables of the prediction model were selected based on physical mechanisms. The main conclusion is that XGBoost demonstrates the highest prediction accuracy among the 22 algorithms, with a mean absolute percentage error of 0.22%.

Keywords heating building, mean room temperature, prediction, machine learning, optimization algorithms