Power Forecasting of Wind Turbines Based on Multi-Input Predictors and Wind Speed Variation Trends

Wind power forecasting plays an important role in the dispatching, operation and monitoring of power grids. In order to solve the problem of low accuracy of existing forecasting algorithms, a novel model using multi-input predictors is presented. First, this paper analyzes the real-world wind speed-power data of a wind farm and finds that wind speed variation tendency is closely related to the output power. Aiming at this discovery, a model which considers multi-step wind speed and its variation trends is proposed. Furthermore, recursive least squares (RLS) algorithm and back propagation (BP) neural network are utilized to obtain parametric and non-parametric models, respectively. Finally, in order to test the feasibility and accuracy of the proposed model, experiments are carried out using the data set obtained from an actual wind farm. The results demonstrate that the proposed model provides a more accurate forecast compared to conventional methods.