Prediction of outdoor air temperature and humidity using Xgboost

Abstract

Outdoor temperature and humidity prediction plays an important role in HVAC intelligent control with huge energy saving potential and identification of the urban heat island effect. The aim of this paper is to develop and evaluate a Xgboost model for the prediction of outdoor air temperature and humidity using acquired data from Shenzhen, China. Datasets fetched from sensor real-time collection and meteorological station interface are utilized as observations, through the construction of Xgboost to predict outdoor temperature and humidity in predictive horizon of 1-3 hours. The effectiveness of the prediction is verified by comparing the prediction with the measured outdoor air temperature and humidity. In addition, statistical tools such as R2 and root mean square error (rmse) are applied to evaluate the performance of the model. The results show the excellent performance of Xgboost in accurately predicting outdoor temperature and humidity by comparison between the measured and predicted outdoor air temperature (R2 > 0.73, rmse < 1.77) and air humidity (R2 > 0.81, rmse < 6.33).