A. M. Filippidou, Statistical and stochastic analysis of hourly wind speed for the simulation of wind energy generation. Application in wind stations in Greece., Diploma thesis, 215 pages, Department of Water Resources and Environmental Engineering – National Technical University of Athens, Athens, July 2015.
Statistical and stochastic analysis of hourly wind speed for the simulation of wind energy generation. Application in wind stations in Greece. Hourly wind speed is examined in wind stations of Central Greece. Statistical analysis of the data includes the presentation of histograms and boxplots, as well as the calculation of the hourly mean values, the standard deviation, the skewness coefficient, the autocorrelation coefficient of first order and the probability calm percentage. Fitting of a theoretical distribution to the hourly wind data is examined, using L-moments. It is concluded that hourly wind speed appears to have double cyclostationarity, because hourly wind speed values vary significantly during the day and according to the season. What is more, due to the significant probability calm percentage, hourly wind speed has to be considered as a mixed type variable. It is divided into the discrete part, which includes the zero values of hourly wind speed and the continuous part, with the positive values only. Stochastic analysis includes the above assumptions. A synthetic timeseries of 1 000 years is generated, which simulates the positive values of hourly wind speed, from the model CAR1. It refers to a double cyclostationary autoregressive model of order 1, while maintaining the skewness. Zero wind speed values are simulated by a first order Markov chain of 1 000 years, which generates the value 0 in case of calm weather and the value 1 in case of positive wind speed, always according to the probability transition matrices. These two timeseries are multiplied so as to result to the final synthetic timeseries of hourly wind speed. The statistical characteristics approach satisfyingly those of the historic data. Finally, wind energy generation is calculated, using the power curve of a specific wind turbine.