Weighted aggregated ensemble model for energy demand management of buildings

Abstract

Accurate building energy consumption prediction is essential for achieving energy savings and boosting the HVAC system's efficiency of operations. Therefore, in this work, a novel ensemble predictive model, which combines the weighted linear aggregation of Gaussian process regression (GPR) and least squared boosted regression trees (LSB), leading to WGPRLSB, is proposed for the accurate estimation of energy usage in the cases of Heating Load (HL) and Cooling Load (CL). Marine predator optimization (MPO) is used to evaluate the optimal values of the design parameters of the proposed methodology. Further, predictive models based on linear regression (LR), support vector regression (SVR), multilayer perceptron neural network (MLPNN), decision tree (DT), and generalized additive model (GAM) are also designed for comparison purposes. The results reveal that the value of RMSE is reduced by 12.4%–70.7% (HL) and 39.7%–64.9% (CL) for WGPRLSB in comparison to the other predictive models. The results of the performance index (PI) also confirm the effectiveness of the proposed model energy consumption prediction for HL and CL. Furthermore, the performance investigation on the second dataset reveals that WGPRLSB achieves the highest value of VAF (97.20%) compared to other designed models. It may be concluded that the proposed WGPRLSB accurately forecasts building energy demands.