A missing data treatment method for photovoltaic installations

摘要

Due to the high installation rate of Photovoltaics (PV) systems, the challenging task of data processing arises. As the number of intermittent PV systems grows especially in the distribution networks, this task becomes even more important. In many cases, the data present inconsistencies, i.e. missing, incomplete or profoundly wrong values. The present paper proposes a method built on machine learning algorithms for missing data completion. The original incomplete PV generation time series are filled and restored with low error. The proposed method can be easily applied in real installations with high rate of data collection and storage needs.