Stochastic modeling of intraday photovoltaic power generation

Renewable energies play an increasing role in power generation worldwide. Electricity generated by photovoltaic power plants is an important factor here. The fact that no solar power is generated at night makes modeling for high resolution difficult. Previous work has therefore been limited to daily variation. However, this obviously leads to a lack in description of the data, a gap which we will fill in this work. To do this, first we filter a cloud cover component from the infeed data by using physical relationships. This variable incorporates the complete stochastic and can be modeled as a non-linear continuous-time autoregression as defined by Brockwell and Hyndman (1992). We fit our model to infeed data in Germany and show that it describes the data better than other comparable approaches. The model enables pricing of derivatives, which is illustrated by a new future contract. This product allows the volume risk of photovoltaic power plants to be hedged. (C) 2019 Elsevier B.V. All rights reserved.