jaue2016053: An Ensemble of BP and Elman Neural Networks for Hourly Power Consumption Forecasting of Building

Abstract

In recent year, Building Distributed Energy System (BDES) has obtained widely application in various buildings. Cooling, heating and power load forecasting play a crucial part in BDES design and operational control. To improve the forecasting accuracy, this paper presents an ensemble of BP and Elman neural networks for hourly power consumption forecasting of building. The BP is designed and trained using feed forward back propagation network while Elman using feed forward network. Therefore, it is helpful for BDES to distribute its electric load optimally, accurately and legitimately. The result indicates that the ensemble network gains lower values in mean square error (MSE) and mean absolute error (MAE) and higher value in relative coefficient (r) compared with individual BP and Elman network. The ensemble model performs acceptable high accuracy that can be used to forecast the hourly power consumption successfully.