用切氏多项式作月降水量的双向时间序列预报试验

TEST ON DOUBLE DIRECTIONAL TIME SERIES FORECASTING MONTHLY RAINFALL BY USING CHEBYSHEV POLYNOMIALS

摘要:

利用文献[1]提出的切比雪夫多项式时间序列预报方法,对月降水量作同年逐月预报和逐年同月预报,将双向时间序列预报加权集成,构成最终最佳预报值。文中讨论了参数的选取和预报效果。对于理想初值的取法制作了详细表格,经数值试验表明,这种双向切氏多项式预报在一定程度上比单向预报提高了精度。

关键词:切比雪夫多项式;时间序列预报;预报集成

Abstract:

In this paper,a double direction forecasting technique for monthly rainfall is presented by means of the iterative algorithm based on Chebyshev Polynomials. The double sirection consists of both inter - monthly rainfall and interannual rainfall. The optimum values of the prediction method are a forecast ensemble based on the two-direction prediction values. Choice of the parameters and their forecasting effects are discussed for Chebyshev Polynomial forecasts. Particularly,in order to select ideal initial values a detailed list is made, and by numerical tests the double direction forecasting technique based on Chebyshev Polynomials proves to have higher precision than the single direction forecasting technique under certain conditions.

Key words:Chebyshev Polynomials;time series forecast;forecast ensemble