Modeling Wind Turbine Power Curve in Complex Terrain: An Efficient Approach Using Big Data and Machine Learning

The wind turbine power curve (WTPC) is a main index to evaluate operating status of wind turbines and the wind energy resource. Traditional WTPC models are unsuitable for wind turbine in complex terrain because they are difficult to reconcile in ensuring model efficiency and improve model accuracy. An efficient approach using big data and machine learning is proposed to model the WTPC. In this paper, the data partition and the SVM algorithm are used to model accurately the WTPC with less computational cost. After eliminating abnormal data, the data are separated into subsets according to the turbulence intensity. Each subset is separated into equal size parts then the centroid of the data in each part is calculated. The centroids are regarded as training data of the support vector machine (SVM) algorithm to model the WTPC. The results show that the approach considering the turbulence intensity and combining parametric with nonparametric model can improve efficiency and accuracy of the WTPC model.