Influence of data dividing of input term on wind speed forecast

摘要

In this study, the wind speed forecast model in which the data in data input term are divided and given is proposed, and optimum sub-input term and the number of sub-input are investigated. Omaezaki, Shizuoka Prefecture, is selected as the target site, and mean wind speed from the present time to 3, 6, or 12 hours later are forecasted by Multiple Regression Analysis. Wind speed and atmospheric pressure data in 1991 to 1999 are used in forecast model. It is found that (i) the forecast error ratio is decreased by dividing the input data, (2) the optimal sub-input term is 3 hour and the number of sub-input is 2 in any forecast term.