An observational method for determining daily and regional photovoltaic solar energy statistics

This paper presents a classical estimation problem for calculating the energy generated by photovoltaic solar energy systems, on a daily, annual, regional and national basis. Our methodology relies on two data sources: PVOutput, an online portal with solar energy production measurements, and modelled irradiance data available for large parts of Africa and Europe, from the Royal Netherlands Meteorological Institute. Combining these, we obtain probability functions of observing energy production, given the irradiation. These are applied to a PV systems database, using Monte Carlo sampling, allowing daily and annual solar energy production to be calculated. These are, in turn, used to calculate solar energy production per municipality. As a case study, we apply this methodology to one country in particular, namely the Netherlands. By examining the variation in our estimates as a result of taking different subsets of PVOutput systems with certain specifications such as azimuth, tilt and inverter loading ratio, we obtain specific annual energy yields in the range of 877-946kWh/kWp and 838-899kWh/kWp for 2016 and 2017 respectively. The current method used at Statistics Netherlands assumes this to be 875kWh/kWp, irrespective of irradiation, meaning the yields were underestimated in 2016 and overestimated in 2017. In the case of the Netherlands, this research demonstrates that an irradiation based measure of solar energy generation is necessary. More generally, this research shows that different types of open data sources may be combined to develop models that calculate the energy production of PV system populations