ABSTRACT
This chapter discusses various methods for power curve modeling and analysis. The importance of power curve modeling goes without saying. The industrial practice of estimating the power curve relies on a non-parametric approach, known as the binning method, recommended by the International Electrotechnical Commission. The power curve established under the free sector has a poor predictive capability for wind power production under general wind conditions. A power curve model should characterize not only the nonlinear effects of wind speed and wind direction, but also the interaction effects among the environmental factors. Kernel regression or kernel density estimation methods have been used for modeling power curves. Additive-multiplicative kernel goes beyond the plain version of kernel density estimation or a kernel regression and employs a special model structure that allows it to handle the multi-dimensional inputs in power curve modeling. Addressing the multi-dimensional power curve problem is essentially a regression problem.
