Review of power curve modelling for wind turbines

• Published in: Renewable & sustainable energy reviews Vol. 21; pp. 572 - 581
• Main Authors: Carrillo, C., Obando Montaño, A.F., Cidrás, J., Díaz-Dorado, E.
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 01.05.2013; Elsevier
• Subjects: Applied sciences; Curve fitting; Energy; equations; Exact sciences and technology; Natural energy; Power curve; Variable speed wind turbine generator; Wind energy; wind power; wind speed; wind turbines; VSWT; Wind energy; WTG; Variable speed wind turbine generator; Power curve; Curve fitting
• ISSN: 1364-0321; 1879-0690
• DOI: 10.1016/j.rser.2013.01.012

Currently, variable speed wind turbine generators (VSWTs) are the type of wind turbines most widely installed. For wind energy studies, they are usually modelled by means the approximation of the manufacturer power curve using a generic equation. In literature, several expressions to do this approximation can be found; nevertheless, there is not much information about which is the most appropriate to represent the energy produced by a VSWT. For this reason, in this paper, it is carried out a review of the equations commonly used to represent the power curves of VSWTs: polynomial power curve, exponential power curve, cubic power curve and approximate cubic power curve. They have been compared to manufacturer power curves by using the coefficients of determination, as fitness indicators, and by using the estimation of energy production. Data gathered from nearly 200 commercial VSWTs, ranging from 225 to 7500kW, has been used for this analysis. Results of the analysis presented in the paper show that exponential and cubic approximations give the higher R2 values and the lower error in energy estimation. With the approximate cubic power curve quite high values of R2 and low errors in energy estimation are achieved, which makes this kind of approximation very interesting due to its simplicity. Finally, the polynomial power curve shows the worst results mainly due to its sensitivity to the data given by the manufacturer.