Cloud Cover NowCasting for Grid Operations

Cloud nowcasting is the process of predicting the movement of clouds and behavior in real-time. This involves utilizing satellite images, along with Numerical Weather Predictions to make short-term forecasts of cloud coverage, precipitation, and other cloud-related phenomena. Cloud cover nowcasting aims at forecasting the position of clouds on a short time scale of up to 0-3 hours ahead. The goal of cloud nowcasting is to provide quick and accurate information about changing cloud patterns which can be used in a variety of applications, including renewable energy, and weather prediction.

REConnect Cloud Nowcasting, leverages the latest Deep Learning technology to provide real-time predictions of cloud behavior in India. Our system combines data from low-resolution remote sensing satellites located in Europe as well as data from INSAT-3DR with outputs from physical models to create highly accurate, region-specific predictions. To ensure our predictions remain relevant and up-to-date, our models undergo continuous training and validation, incorporating on-ground GHI data from various locations across India. This approach considers the seasonal and regional variability of cloud patterns in India and ensures the accuracy of the models.

Traditional approaches of short-term solar forecasting include cloud tracking with optical flow using ground based sky imagery, while recent developments focus more on using remote sensing satellite data as starting points, and use atmospheric motion vectors for extrapolation. However, these methods are heavy on computation time and take multiple hours when applied to larger areas and fail to predict possible changes in shape and size of the clouds. 

For short term nowcasting to be effective, the computation has to be fast enough that the results can be converted to use in real world application. Longer computation times make the initial time steps in the forecast useless. 

This is where AI models with deep architectures are found to be beneficial. 

The main application of Cloud nowcasting is  to optimize the energy generation of solar energy by providing real-time information about cloud cover, which affects the availability of solar radiation. It helps the renewable generators in better grid management by providing real-time and accurate information about the location and movement of clouds, which affect the amount of solar radiation reaching the ground. Additionally it can also help grid operators to optimize energy storage and reduce cost by enabling them to predict periods of excess production and store it for later use.