Wind Turbine Output Power Prediction in a Probabilistic Framework Based on Fuzzy Intervals

Abstract

Wind turbine as a popular growing renewable energy source has found its real location in the modern smart grids more than ever. The useful and efficient deployment of wind turbine in the smart grids depends on accurate modeling of its behavior including the amount of output active power and its continuous variations during the time. Such a complex high-dependent model requires accurate prediction models which can not only follow the wind turbine output power, but can also model its uncertainty effects in the prediction window time. This is mostly triggered due to the extremely instable and casual nature of wind behavior other than the non-linear dependency of wind power on its speed. This paper develops a probabilistic framework for predicting the wind turbine output power based on interval concept. The proposed model can properly predict the wind forecast samples when minimizing the prediction bandwidth and thus increasing the information contained. This is achieved using a lower and upper bound scheme constructed without any prior judgment regarding the uncertainty of prediction distribution. This makes the proposed model so capable of proving realistic intervals which can help the smart grid operator in the optimal scheduling process. To improve the presentation of the proposed model, an efficient search system based on social spider optimization is developed to optimize the prediction model adjusting parameters. The highly trustable and accurate performance of the proposed model is examined using the experimental wind turbine dataset recorded in two different wind turbines in the Australia coastal areas. The results show the promising and accurate prediction results found by the proposed model.