气温和降水时变影响参数的多步预测模型

本文提出了一种气温和降水时变影响参数的多步预测建模新方案。该方案用时间序列均生函数外延矩阵生成的主分量作基函数对序变量建模，同时考虑变量前期数据对后期数据的影响，在模型中引入一个时变影响参数Ｋ１。并用方差分析法求出Ｋ１的显著周期和未来时刻的Ｋ１＋１值，从而建立气温和降水序列的时变影响参数的多步预测模型。该模型用于四川省２０个地、市的气温和降水预报数值试验，其历史拟合率和试报结果与实况值的比较表明，

MULTI-STEP FORECASTING MODEL OF TEMPERATURE AND PRECIPITATION BASED ON THE TIME-VARYING INFLUENCE PARAMETER

In this paper, a new multi-step forecasting model of temperature and precipitation series based on the time-varying influence parameter is presented. The approach is based on to consider the influence of the earlier stage values on later stage values. Time-varying parameter Kt is introduced into the model of principal component generated by mean generating function of temporal sequence. The multi-step forecasting model of temperature and precipitation series is developed by the time-varying parameter values Kj+1 calculated with significance periods of K. The models are used to multi-step forecasting of temperature and precipitation in 20 regions of Sichuan. The comparison of fitting test and forecasting results with real values show that this model can be used in multi-step forecasting of time series variate.