Abstract
Building integrated photovoltaics (BIPVs) have become increasingly popular in urban areas due to their potential to provide zero-emission energy in buildings. BIPVs enable the use of MO of photovoltaic (PV) modules, resulting in multiple generation peaks throughout the day, which can closely match the consumption profile of a building. However, designing a PV output power forecasting model for BIPVs can be complex, as multiple generation peaks may occur. In general, prior works focused on the design of a PV output power forecasting model and the study of the impact of tilt and azimuth angle variations on the annual yield of a PV system independently. The interaction between the two problems has received little or no attention in the literature. Considering the recent rapid development of BIPVs, it is therefore extremely important to understand this. This article, therefore, aims to quantify the impact of mixed orientations on the accuracy of a PV output prediction model for an 181.15 kWp BIPV system located in Trondheim, Norway. The results indicate that the mixed orientations resulted in a notable rise in forecast error, with a maximum increase of 51% in root mean squared error, when compared with a baseline model that accounted for all orientations. These findings have important implications for developing practical predictive energy management systems for BIPVs.
