“…It has been extensively used in analysis of signal together with pattern recognition (see [5]), statistical computation (see [9]), and computational fluid dynamics (see [29]). In recent years, it also has been successfully applied to the order-reduction for the Galerkin method (see, e.g., [10,11]), the finite element method (see, e.g., [18,22]), the FD scheme (see, e.g., [24,31]), finite volume element method (see, e.g., [21,23]), and reduced basis methods (see, e.g., [1,6,28]) for PDEs. However, the most existing POD reduced order methods (see, e.g., [1, 2, 5-7, 9-11, 18, 21-25, 28-31]) are built by the POD basis formed with the classical solutions at the all time nodes on [0, T ], before repetitively computing the reduced order solutions at the same time nodes.…”