다양한 유사성 척도를 이용한 바람권역의 군집분석

Abstract

To assess wind resources in a given area, it is necessary to select a representative measurement point by identifying a wind zone where the wind resource characteristics are homogeneous. In this study, to identify the spatial homogeneity of wind resources, it was necessary to test several similarity measures for clustering analysis, such as time-series wind vector similarity, Pearson's correlation coefficient of time-series wind speed, the cosine distance of time-series wind direction, the index of agreement of time-series wind speed, the 24-hour autocorrelation function, and the principal components of wind resource factors. It was found that the primary components of wind resources were the Weibull scale and shape factors of wind speed distribution, while terrain elevation and the 24-hour autocorrelation function were chosen as the secondary components. The similarity measures were applied to Jeju Island, which has a simple topography, and the Pohang region, which has a complex mountainous topography, to classify the wind zones; while poly-serial correlation coefficients, together with box plots, were used to evaluate whether there was a significant statistical difference in the wind resource factors within a given cluster. In conclusion, it was confirmed that the time-series wind vector similarity and the primary components of wind resource factors are the most effective similarity measures of clustering analysis.