While solar energy has definite advantages over conventional power sources such as oil and coal, its unpredictable nature poses a number of potential problems for the electrical grid. With conventional generators, grid operators can relatively easily increase production to meet demand. However, the unpredictable nature of solar energy can make it difficult to properly balance production and demand which can lead to grid instability. Accurate solar irradiance forecasts can reduce the unpredictability and help to ensure a stable grid.
In this paper1 we present a method for short-term solar irradiance forecasting using gridded global horizontal irradiance (GHI) data, estimated from satellite images. We use this data to first create a simple linear regression model with a single predictor variable. We then discuss various methods to extend and improve the model. We found that adding predictor variables and partitioning the data to create multiple models both reduced prediction errors under certain circumstances. However, both these techniques were outperformed by applying a data transformation before training the linear regression model.
We also discuss a set of methods for identifying “rare weather events” by discretizing Global Forecast System (GFS) data by transforming the data using either principal component analysis or a discrete wavelet transform, then discarding and round the transformed data. While the current work only investigates identifying past events, future work could investigate predicting these events and using those predictions to improve solar irradiance forecasts.