Short-Term Forecasting of Wind Speed and Power - A Clustering Approach

摘要

In this study, we develop a mixed ARMA model that incorporates the wind direction into short term wind speed and wind power output forecasts. For this purpose, existing association between the wind speed and wind direction are examined using a clustering approach. Using k-means algorithm, wind directions are classified based on the accompanying wind speeds. Using an ARMA model for forecasting the wind direction, those values are associated with the formed clusters by using dummy variables. These dummy variables are employed in the mixed ARMA model. The analysis indicates that incorporating wind direction provides slightly but consistently better estimates for the wind speed for short term forecasts. Improvements in forecasting accuracy for the wind power output are also realized by employing mixed-ARMA models.