Preliminary Analysis of Short-term Solar Irradiance Forecasting by using Total-sky Imager and Convolutional Neural Network

Abstract

The installation of photovoltaic system (PV) is increasing rapidly across the world. However, the fluctuation of PV output causes serious challenges in the power grid operation. Among the fluctuation, a quick part of fluctuation is mainly caused by the change of cloud coverage. A total-sky imager (TSI), measuring device to take sky and cloud, could be useful for short-term solar irradiance forecasting. In this paper, a convolutional neural network (CNN) is applied to forecasting model (called CNN model) to forecast 5-20 min ahead of global horizontal irradiance (GHI) using total-sky images and lagged GHI. To verify the effectiveness of CNN, three forecasting models are compared. They are the persistence model, the CNN model using only total-sky images, and the CNN model using both total-sky images and lagged GHI. From the computation, the proposed CNN model using both total-sky images and lagged GHI performs root-mean-square error (RMSE) of 49-177W/m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> , 93-146W/m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> , 71-118W/m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> in sunny day, partly cloudy day and overcast day, respectively. From these results, the proposed method is shown to be suitable for short-term solar irradiance forecasting.