Combined PV Power and Load Prediction for Building-Level Energy Management Applications

Abstract

In order to successfully integrate renewable energy technologies, the requirements of local energy management systems are becoming increasingly complex, as is the sector integration of electricity, heat and transportation. To address this, this study investigated the combination of machine learning-based PV power and load demand prediction approaches to forecast residual load at the building level. The forecast accuracy, seasonal dependencies and the effects of single forecasts on the residual load were evaluated by means of three different metrics, namely: mean absolute error (MAE), root-mean-square error (RMSE) and the mean absolute scaled error (MASE). The applicability of the combined forecast was tested via a case study of integrated battery-electric vehicles and a PV system in an existing commercial building. The results show how the residual load forecast can help schedule grid-friendly charging demand and optimize PV self-consumption.