Exogenous Parameters in Solar Forecasting

The ability to predict solar radiation reliably is crucial in optimizing solar energy integration, ensuring grid stability and regulating energy markets. One way to improve accuracy in forecasting solar radiation with time series modeling is to use exogenous variables (e.g. temperature, humidity, pressure, wind speed, and direction) in addition to solar radiation measurements. Evidence from existing studies indicates that the extent to which such exogenous variables can improve solar forecasting is largely dependent on the type of algorithm used. Our results indicate that the scope of the prediction target (lag duration, number of steps ahead) also plays an important role in determining the ability of exogenous variables to improve solar forecasting results. More specifically, the accurate pairing of exogenous variables and forecasting algorithms can help achieve accuracy improvements with longer lags at diverse horizons. These results argue in favor of a multi-modeling approach where specific forecasting configurations are determined dynamically for each choice of time series input.