A time series-based approach for renewable energy modeling

Despite the growing literature on renewable energy sources, causal relationships between the variables that are selected as inputs of the models proposed in forecasting studies have not been investigated so far. In this paper, a novel approach to decide prediction input variables of wind and/or temperature forecasting models is suggested. This approach uses time series techniques; more specifically, Granger causality and impulse-response analyses between some meteorological variables. To conduct our study, wind speed, temperature and pressure data obtained from different regions of Turkey are employed. The results suggest that bidirectional causal relationships exist between these variables and that short-run dynamics differ with respect to location (inland versus coastal area). From this, it is concluded that renewable energy models must be built accordingly to improve prediction accuracy.