Modeling of wind pattern and its application in wind speed forecasting

Abstract

As wind power penetrations increase dramatically, wind power forecasting is increasingly becoming one of the fundamental strategies to coordinate wind generators together with thermal or other traditional generators in hybrid power systems. This paper focuses on very short term wind speed forecast based on historical wind speed data. The term wind pattern is proposed in this paper to characterize one type of trend of weather process and duration of each pattern may be different. It is used in the forecasting process to depict the possible trend of wind speed variation, which is beyond the description of most time series methods. When combined with time series methods such as ARMA, wind patterns are expected to improve forecast results. In this paper, parameters that identify wind speed pattern are defined and wind pattern model is established. An ensemble forecast method of ARMA and wind pattern is presented. Wind speed data of a wind farm in Inner Mongolia, which lies in the northern part of China, is used to verify the proposed method. The new forecast method with wind pattern model shows better results in the very short term wind speed forecast.