| 초록 |
In the prediction error identification method (PEM) for the auto-regressive, moving average, exogeneous input (ARMAX) model, the optimization problem has a useful feature that the optimal parameters corresponding to the auto-regressive and exogeneous input part can be solved analytically. Using the feature, we can reduce unknown parameters subject to nonlinear iteration in PEM and propose a modified prediction error identification method (MPEM). It shows better robustness than PEM due to the reduction of the searching space in the iterative optimization step and removes the possibility of setting poor initial values for the auto-regressive and exogeneous input part. |
