24-Hour photovoltaic generation forecasting using combined very-short-term and short-term multivariate time series model

In order to achieve greenhouse gas reduction and renewable energy penetration target, photovoltaic generation can play an important role as an alternative to fossil fuel based generation in South Korea. However, due to its variability and uncertainty it is required to develop the model to forecast PV generation as accurately as possible. In this paper, a combined very-short-term and short-term model for 24-hour generation forecasting is proposed. Firstly, after considering weather factors affecting PV generation at the sample site in South Korea, the best single model for each time horizon is selected by the least forecasting error. Secondly, those are combined by the optimal value of forecasting time. As a result, the forecasting accuracy of the combined model is improved more than a single model for 24-hour generation forecasting.