Applications of Gaussian Process Models in Regression Analyses and Stochastic Simulations of Wind Speed Data

• Main Authors: Wei-Chih Hsu, 徐偉誌
• Other Authors: Jian-Ye Ching
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/54467348074216487761

博士 === 國立臺灣科技大學 === 營建工程系 === 96 === Wind speed prediction and simulation are ardent topics all the time because of its stochastic properties and significance in wind engineering. Several numerical techniques, e.g. auto-regressive moving average model, artificial intelligence technique etc., were developed for solving the related problems in recent years. In this research, a probabilistic model, named Gaussian process model, is proposed to consider the uncertainties of wind speed. Moreover, Bayesian analysis and transitional Markov Chain Monte Carlo method are employed to find the model hyper-parameters. Three examples for different issues are presented to demonstrate its practicability and satisfactory interpolation performance. Furthermore, the results also show that the various statistic properties, including exceedance probability, data correlation and distribution, of simulated wind speed are consistent with them of the training wind speed data.