PV Predictions Made Easy: Flexibility Through Simplicity

Accurate predictions of PV output power play an important role in supporting the energy transition. This article presents an approach that aims at such predictions for PV installations on a household level. It is designed to be implemented easily on home energy management systems with low computational power. The presented prediction algorithm is self-learning, does not need physical parameters of the PV installation and can deal with changing circumstances such as objects (partially) blocking the PV panels. The method is straightforward to implement and a reference implementation in Matlab is given. An evaluation demonstrates that when the inputs used for the approach (irradiance) are correctly predicted, the predicted PV power output is also accurate.