Towards a tuning method of PV power measurements to balance systematic influences

With rapid deployment and penetration rates of residential photovoltaic (PV) systems in the distribution grid, there is growing need for accurate assessment of the real-time power generation for grid management and energy market operations. Many of these installed PV systems report their live power generation to online databases and can be used as references to estimate the power generation of neighbouring systems. Upscaling approaches have demonstrated their capability of using the data from these reference PV systems to estimate the power output of target PV systems that do not report their power generation data. However, there is an inherent issue with the representativeness of these reference PV systems power data, e.g. due to quality issues or system specific influences such as shading. Three methods were developed by the authors in earlier work: (1) a parametrisation of PV system metadata and quality control of the measured power, (2) a tuning routine that detects diurnal influences from shading and tunes the PV power in order to reach the expected generation without any shading. And (3) a method which eliminates high variances in k_pv based upscaling. An extensive cross-validation with 308 systems in Canberra, Australia in this paper shows significant improvements as a direct result of the application of these three methods. Furthermore, we present the preliminary findings for developments in: the parametrisation of shaded/multi-azimuth reference PV systems, as well as a method to reduce inertia in the shade detection and tuning. Overall, we successfully improve the management of reference PV system power data for use in upscaling.