Abstract of Thesis presented at COPPE/UFRJ as a partial fulfillment of the requirements for the degree of Master of Science (M.Sc.)
Wind Time Series Clustering for Estimation of Wind Farms Generation Availability
Tatiana Araújo de Souza
March/2008
| Advisors: |
Alexandre Pinto Alves da Silva
Carmen Lucia Tancredo Borges
|
| Department: |
Eletrical Engineering |
This work has the purpose to identify clustering techniques that can reduce the number of states of wind which are necessary for modelling the behavior of the wind, without loosing the representability of the original series, in order to decrease the computational eŽort in reliability studies of wind farms. The reduction of the states of wind can be obtained by using clustering techniques, such as K-Means and Fuzzy C-Means. First, the univariate clustering is done, which means that the original series is used as input of the algorithm. Second, the original series is divided into subseries, each one containing 6 values, and the matrix formed by these subseries is the input of the algorithm, so that the multivariate clustering can be done. The goal of this clustering is to catch the behaviour of the wind in one hour period. Several numbers of clusterings are tested for some wind time series collected in diŽerent places of Brazil. Both the results of univariate and multivariate clusterings are used in wind farms reliability studies, for which diŽerent reliability indexes are calculated (IWE, IWP, EAWE, EGWE e WGAF). The results of the reliability indexes for these new series obtained by clustering techniques are compared with original serie's reliability indexes in order to evaluate the quality of these partitions.
