A novel ultra-short-term wind speed prediction method based on dynamic adaptive continued fraction

Accurate and fast wind speed prediction can enhance the control performance of wind turbines and promote the dispatch of wind power integration. However, the intermittent and fluctuating characteristics of wind speed impose challenges on prediction models. To establish a high-accuracy and fast wind speed prediction model, a novel dynamic adaptive continued fraction (DACF) method is proposed. A more universally structured continued fraction model is utilized to capture the variation pattern of wind speed series. To improve both accuracy and computational efficiency, the stepwise parameter estimation method is introduced to achieve the dynamic and adaptive estimation to the structure parameters of continued fraction model. Moreover, considering the fluctuation feature of wind series, the predicted results of continued fraction model are corrected by the fluctuation residual correction function. And the convergence of the proposed model is theoretically proved. The proposed model is verified from four case studies based on two wind datasets with different time scales of two wind farms and achieves considerable accuracy improvement with 5.37%-28.34% reduction of root mean square error compared with contrasted models. The experimental outcomes confirm that the proposed model outperforms six contrasted models in terms of prediction precision and modeling time.