A Statistical Analysis of Transformation Methods for Wind Power Curve Modelling

The dissemination of wind speed is the foundation for the calculation of wind energy potential necessary for the design of wind farms. Thus, a perfect determination of the probability distribution of wind speed is an imperative parameter to measure before estimating the wind energy potential over a particular region. In this paper 10 different distributions have been compared to fit the wind speed data. The goodness of fit of the wind speed was analysed by Kolmogorov-Smirnov test, Anderson Darling test and Chi-square test. To avoid the practical difficulties in obtaining the wind power data of different stations, two transformation methods have been proposed to calculate the wind power from the observed wind speed data of the particular station. This paper describes and compares the wind power modelling using two different transformation methods namely four parameter logistic and five parameter logistic power expressions. The constants of these expressions are evaluated using the Differential Evolution (DE) and Particle Swarm Optimization (PSO) algorithms. The powers obtained by the two transformations are compared with empirical power, data and their performance analysed using the error metrics MAD & RMSE. The results indicate that the five parameter logistic (DE) transformation method is the better method to evaluate the power from the observed wind speed data.