METHOD FOR PREDICTING NONLINEAR SEASONAL TIME SERIES

摘要

PROBLEM TO BE SOLVED: To provide a method of predicting a future value of a time series from a seasonal pattern and the past value of the time series having a nonlinear relation between the past value and future value.

SOLUTION: A short term electric power demand is predicted by applying a dynamic modeling of the seasonal time series data using an exponential smoothing technique with kernels. The kernels are used to predict the future values of the time series from the past value by applying a nonlinear functions such as least-squares radial basis function regression or support vector regression using a Gaussian kernel.

COPYRIGHT: (C)2011,JPO&INPIT