Improved satellite-derived PV power nowcasting using real-time power data from reference PV systems

Rapid growth in the global penetration of solar photovoltaic (PV) systems means electricity network operators and electricity generators alike are increasingly concerned with the short-term solar forecasting (nowcasting) of solar irradiance. This paper proposes a methodology that considers a varying number of available reference PV systems for supporting satellite-derived PV power real-time nowcasting. We evaluate conventional satellite-only and upscaling-only PV fleet estimate methodologies and compare them to two newly developed correction and hybrid cases. When using only a single reference PV system to estimate the aggregated power of 48 independent target PV systems for the location of Canberra, Australia; we show that the newly proposed correction or hybrid cases improve the performance of the satellite-derived PV power estimate medians in terms of MBE, rMBE, RMSE and rRMSE from 0.031 W/W_p,7.46%, 0.079 W/W_p and 23.4%, down to 0.006 W/W_p,-0.711%, 0.068 W/W_p and 20.0%, representing relative improvements of 80.6%, 90.5%, 13.9% and 14.5%, respectively. Similarly, when using 30 reference PV systems, we report median improvements from 0.036 W/W_p,8.25%, 0.083 W/W_p and 24.8%, down to 0.01 W/W_p,1.41%, 0.049 W/W_p and 11.4%, representing relative improvements of 72.2%, 82.9%, 41.0% and 54.0%, respectively. We discuss the fundamental challenges facing the use of reference PV systems, satellite-derived power estimates, combining the two data sources, and the knowledge required to address these issues. We ultimately conclude that combining satellite-based PV power estimates with data from reference PV systems is always more beneficial than either on their own.