为了研究径流的长期演化模式及自记忆特征,采用自相关函数确定重构序列的回溯阶数,并构建基于经验模态函数(EMD)分解的灰色自记忆GM(1,N)预测模型。分析回溯阶和自相关函数之间关系,进一步阐明自记忆原理在水文领域应用的合理性。结果表明,黄河花园口86年天然径流序列存在长期演化的自记忆特征,演化模式可由7个内在本征模态函数(IMF)和趋势项组成;受流域水资源开发及气候变化因素影响,1965年后,径流演化模式发生变化,出现周期衰减现象;回溯阶与时间序列自相关函数变化相对应,序列平稳化后,依据自相关函数确定回溯阶数,可提高自记忆模型预测精度。
To better understand the evolution modes and self-memory characteristics of the Yellow River annual runoff,a gray self-memory prediction model is developed by using empirical mode decomposition theory.A auto-correlation function is adopted to define the backdate orders of this model to improve its accuracy.The results show that the runoff series at Huayuankou gauge can be decomposed into seven intrinsic mode functions(IMF) and a tendency function.Since 1965 the evolution modes and periodicity of runoff series has changed under the influences by water resources development and climate change in this river basin.