An Improved Cooling Load Prediction Method for Buildings with the Estimation of Prediction Intervals

Prediction errors are inevitable for all building cooling load prediction methods. Prediction errors are harmful for the building energy management strategies which are based on the cooling load prediction. This study aims in developing a cooling load prediction error correction approach to reduce prediction errors, and proposing a cooling load prediction intervals estimation approach to describe the uncertainties of predicted cooling load quantitatively. Moreover, the distance correlation is used for feature selection. Evaluations are made using a virtual large office building simulated by EnergyPlus. Support vector regression (nonlinear model) and multivariable linear regression (linear model) are employed to evaluate the performance of the proposed error correction approach. The prediction interval coverage probability (PICP) is used to evaluate the performance of proposed prediction intervals estimation approach. Results show that the proposed error correction approach can do improve the prediction accuracy and the proposed prediction intervals estimation approach is reliable. (c) 2017 The Authors. Published by Elsevier Ltd.