Power forecasting of solar photovoltaic power systems based on similar day and M5' model trees

Distributed energy systems with solar photovoltaic (PV) power grow rapidly in recent years in China. Accurate power forecasting is important for deep utilization and new exploitation of PV power. This study advances power forecasting modelling of PV systems using a method based on similar day selection and M5' model trees. A one-day-ahead power output forecast model is derived for an experimental station. The meteorological forecasting data, historical power output, and the M5' model trees are integrated to build this model. A 1-kW amorphous silicon solar photovoltaic power experiment system is used to verify the validity of the proposed method. The test results show that the forecasting model and the method are effective and promising.