In view of the problem of modeling and forecasting for small sample oscillating series, which exists widely in the
real world but to be paid less attention by people, this paper proposed oscillating GM(1,1) power model with system latency
and time-varying parameters. The two ranks’parameters package formula is presented under the least squares criterion.
On this basis, a nonlinear optimization model is employed to seek the best exponent and time interaction parameters, in
order to identify the oscillating characteristics behind raw data. Finally, the proposed model is applied to forecast emergency
resource demand, and the modeling precisions are compared among the traditional GM(1,1) power model, ARIMA and
EMD- ARIMA. The results show that the oscillating GM(1,1) power model has the highest accuracy.