Forecasts of PV power output using power measurements of 80 residential PV installs

Abstract

We describe a new method to forecast the power output from photovoltaic (PV) systems under cloudy skies. We forecast cloud-induced fluctuations in output power by using measurements from 80 residential rooftop PV systems as an input to a forecasting algorithm described here. We compare the performance of our new method to results from our numerical weather model, and also to forecasts based on our images from a ground-based sun-tracking sky camera. Our numerical weather model provides forecasts of irradiance up to several days in advance. In comparison, our network of PV systems can forecast output up to an hour in advance. Our sky-camera image analysis algorithms can be used to forecast 10 minutes in advance. We also show how hybrid methods can improve the accuracy of hour-ahead forecasts.