DNI nowcasting applying a differential approach method into sky camera images

Concentrated solar energy systems, particularly Concentrated Solar Power (CSP), need to capture a significant amount of Direct Normal Irradiance (DNI) and demand high precision in forecasting their availability in the near future. Understanding how DNI intensity fluctuates across its path through the atmosphere down to the Earth's surface is crucial for harnessing solar energy effectively, as atmospheric conditions define the amount of irradiance that can be captured by CSP systems. To enhance operational efficiency, accurate DNI forecasts are required for short, medium, and long-term durations. In this paper, we present a model that automates two independent studies based on previously published papers by the authors and on sky camera images and solarimetric data. These studies estimate DNI attenuation caused by cloudiness using an artificial neural network (ANN) and measure the optical flow in sky images with the Lucas-Kanade method to forecast the future position of clouds. The proposed model aims to predict DNI in the short-term (5, 10, 15, 20, 25, and 30 minutes) for the entire year of 2021, under various sky conditions (overcast, partially cloudy, and cloudless). The results demonstrate an annual correlation coefficient ranging from r = 0.90 to r = 0.87 in the forecasting horizon from 5 to 30 minutes.