基于ARIMA和GA-Elman神经网络的新疆年降水耦合预测研究

Coupling Prediction for Annual Precipitation in Xinjiang Based on ARIMA and GA- Elman Neural Network

【目的】提高降水预报的预测精度,准确预测一个地区未来的降水量,可以提高该地区防灾减灾的能力,更好地为工农业生产生活提供决策参考。【方法】以年降水时间序列为研究对象,利用差分自回归移动平均(ARIMA)和GA-Elman神经网络技术建立一种耦合预测模型。该模型首先根据年降水时间序列建立ARIMA模型,拟合它的线性结构部分,基于原始降水序列和ARIMA模型的预测值、残差序列,利用GA-Elman神经网络技术进行耦合建模。将该模型应用于新疆年降水量的预测预报,并与单一的ARIMA模型、GA-Elman神经网络模型进行比较。【结果】耦合模型的归一化均方误差、平均绝对误差、后验差比值及小误差概率分别为0.287,9.581,0.241和1,均优于ARIMA模型、GA-Elman神经网络模型,预测精度得到了明显的提高,预测精度等级为好。【结论】基于ARIMA和GA-Elman神经网络的耦合预测模型具有更高的预测精度,可用于新疆的年降水量预报。

【Objective】Improving the prediction accuracy of precipitation forecast and making an accurate precipitation forecast for the future can improve the ability of disaster prevention and mitigation of the region,and provide the decision- making reference better to industrial and agricultural production as well as daily life.【Method】By taking time series of annual precipitation as the research object,using difference Autoregressive Integrated Moving Average（ ARIMA） and GA – Elman neural network technology to construct a coupling prediction model. Firstly,the ARIMA model was established according to the annual rainfall time series to fit the linear structure part. Secondly,based on the original precipitation sequence and predicted values,residual series of ARIMA model,the coupling model was constructed by using GA – Elman neural network technology.【Result】The model was applied to predict and forecast annual precipitation in Xinjiang,and was comparedwith a single ARIMA model and a single GA – Elman neural network model. The results showed that the normalized mean square error,mean absolute error,posteriori error ratio and small error probability were 0. 287,9.581,0. 241 and 1,respectively,which were better than those of ARIMA model and GA- Elman neural network model,and the prediction accuracy was improved remarkably,the grade of prediction precision was desirable.【Conclusion】The coupling prediction model based on ARIMA and GA – Elman neural network showed higher prediction accuracy,which ccould be applied in the precipitation prediction and forecast in Xinjiang.