| 摘要: |
| 【目的】研究最优的枯季径流预测模型，为流域水资源管理提供依据。【方法】建立基于差分自回归移动平均（ARIMA）、人工神经网络（ANN）和多元线性回归（MLR）3个单项模型的简单平均组合和最优加权组合预测模型, 并将单项预报模型和组合模型应用到石羊河流域支流西营河的枯季径流预测中，采用相关系数、确定性系数以及均方根误差对各模型预测精度进行比较。【结果】单项预测模型中，仅ARIMA模型通过了确定性系数检验；最优加权组合模型的预测精度较简单平均组合模型高；组合预测模型中，仅ARIMA-MLR和ARIMA-ANN最优加权组合模型的确定性系数高于所有单项预测模型。【结论】最优加权组合模型的精度不但取决于各单项预测模型的精度，也与其之间的相关性有关，适合西营河枯季径流预测的最优加权组合模型是ARIMA-MRL和ARIMA-ANN组合模型。 |
| 关键词: 枯季径流预测 差分自回归移动平均 人工神经网络 多元线性回归 组合预测模型 西营河 |
| DOI： |
| 分类号: |
| 基金项目:国家自然科学基金项目（50879071）；西北农林科技大学青年学术骨干支持计划项目 |
|
| Drought period stream-flow forecasting based on optimal weighted combination model |
|
|
| Abstract: |
| 【Objective】The optimal drought period stream-flow forecasting model is studied to provide a basis for the management of river basin water resources.【Method】This paper presents combined forecasting models of simple average and optimal weighted based on the principle of ARIMA,artificial neural network(ANN) and multiple linear regression(MLR),and applies the single models and the combinated models to the Xiying River of Shiyang River drought period stream-flow forecasting,and the results of the accuracy of those models,compared by using correlation coefficient,deterministic coefficient and root mean squared error,are also presented in this paper.【Result】In single forecasting models,only the ARIMA model passes the examination of deterministic coefficient.The accuracy of optimal weighted combinated model is higher than simple average combination.In the combinated forecasting models,deterministic coefficient of the combination of ARIMA and MLR and the combination of ARIMA and ANN are higher than all single forecasting models.【Conclusion】The accuracy of the optimal weighted model not only depends on the accuracy of the single forecasting models,but also depends on the correlation of the error of single forecasting models.The best optimal weighted combinated models of drought period stream-flow in Xiying River is the combination of ARIMA and MLR and the combination of ARIMA and ANN. |
| Key words: drought period stream-flow forecasting ARIMA ANN MLR combination prediction model Xiying River |
