Sparse Heteroscedastic Multiple Spline Regression Models for Wind Turbine Power Curve Modeling

An accurate wind turbine power curve (WTPC) plays a vital role in forecasting and condition monitoring. There are two major shortcomings of current WTPC models that prevent more estimation, limited nonlinear fitting ability the lack in-depth understanding complex characteristics WTPC. This paper proposes novel regression to overcome these disadvantages simultaneously. First, they make use multiple spline (MSRM) with different basis functions numbers knots describe relationship between speed power. Moreover, sparse prior distributions help avoid adverse effects redundant mapping features useless on model performance. Second, embed heteroscedasticity modeling into MSRM based Gaussian Student's t-distributions, respectively. Finally, heteroscedastic t-distributions will be constructed named as SHMSRM-G SHMSRM-T, We compare proposed fifteen benchmark models, find can generate WTPCs than others seasons farms. Thus, it is important consider together constructing models.