Determination of the wind energy potential using the statistical methods and the genetic algorithm

The increase in population and the inadequacy of available energy resources put human being into the search of alternative energy resources during the human history. Today, the rapid depletion of fossil energy resources and increasing environmental concerns have increased the interest in renewable energy forms. Among the renewable energy forms, the interest in wind energy has increased considerably in recent years. In order to make a wind energy investment in a region, the wind speed in the region must be measured for long term and the wind energy potential of the region must be determined. Wind speed is the most important variable in determining the wind energy potential and the effective modeling of the wind speed is very important for both evaluating the profitability of the investment and choosing the proper turbine types. In the modeling of the wind speed, a large number of distributions are used. In this thesis, Weibull, gamma, normal and log-normal distributions and twocomponent mixture forms of these distributions were compared using wind speed observations measured in three different regions. Model parameters were estimated with the maximum likelihood method. Newton-Raphson (NR), Broyden-Fletcher-GoldfarbShanno, Nelder-Mead and Simulated Annealing (SA) methods were used in obtaining the estimators. As a result of the comparisons, it was determined that the two-component mixture Weibull (WW) distribution was one of the successful distributions. In the next section, the genetic algorithm (GA) method with a search space based on the Expectation Maximization (EM) algorithm and the bootstrap technique has been proposed to obtain the maximum likelihood estimators of the WW distribution, and the proposed method has been compared with the NR, EM and SA methods. According to Kolmogorov-Smirnov test statistics, root mean square of error, coefficient of determination, and deficiency criteria, the proposed GA method provided the most efficient estimates. Keywords: Genetic Algorithms, Finite mixture distributions, Weibull distribution, Expectation Maximization Algorithm, Wind speed modelling