首页|基于EEMD-SE-LSTM组合模型的开都河日径流模拟研究

基于EEMD-SE-LSTM组合模型的开都河日径流模拟研究

Study on daily runoff simulation of Kaidu River based on EEMD-SE-LSTM combined model

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
为提高开都河日径流模拟的精度和更科学地进行开都河水资源的管理与规划,在集成经验模态分解(EEMD)的基础上进行样本熵(SE)重构来完成长短期记忆网络(LSTM)组合模型的构建.采用集成经验模态分解提取开都河日径流序列中具有物理含义的信息,得到一系列本征模态分量(IMF)及一个趋势项(Res),计算每个分量的样本熵,复杂程度接近的子序列叠加为新序列,建立长短期记忆神经网络模型进行预测,叠加得到最终模拟值.结果表明:EEMD-SE-LSTM 组合模型日径流模拟的精度得到提高,其确定系数 R2=0.81、纳什效率系数 NSE=0.73,均高于 LSTM 模型的 R2=0.73、NSE=0.52 和 EEMD-LSTM 模型的 R2=0.64、NSE=0.63;EEMD-SE-LSTM 组合模型的日径流模拟准确性更高,其评价指标(R2=0.81、NSE=0.73)高于其他单一模型 SVM(R2=0.70、NSE=0.58).EEMD-SE-LSTM 组合模型提高了日径流模拟精度,可以更好地为开都河水资源管理与规划提供科学依据.
In order to improve the accuracy of daily runoff simulation of Kaidu River and carry out the management and planning of Kaidu River water resources by science,the sample entropy(SE)reconstruction is carried out on the basis of ensemble empirical mode decomposition(EEMD)to complete the construction of long short-term memory network(LSTM)combination model.The ensemble empirical mode decomposition is used to extract the information with physical meaning in the daily runoff sequence of Kaidu River,and a series of intrinsic mode components(IMF)and a trend term(Res)are obtained.The sample entropy of each component is calculated,and the subsequences with similar complexity are superimposed into a new sequence.The long-term and short-term memory neural network model is established to predict,and the final simulation value is obtained by superposition.The results show that:The accuracy of daily runoff simulation of EEMD-SE-LSTM combined model is improved,and its determination coefficient R2=0.81 and Nash efficiency coefficient NSE=0.73 are higher than those of LSTM model(R2=0.73,NSE=0.52)and EEMD-LSTM model(R2=0.64,NSE=0.63).The daily runoff simulation accuracy of EEMD-SE-LSTM combined model is higher,and its evaluation index(R2=0.81,NSE=0.73)is higher than other single models SVM(R2=0.70,NSE=0.58).The EEMD-SE-LSTM combined model improves the accuracy of daily runoff simulation and can better provide a scientific basis for water resources management and planning of the Kaidu River.

Integrated empirical mode decompositionSample entropyLong short-term memory networkCombined modelDaily runoff simulatio

丁占涛、安杰、吴国洋、宋昱锋、罗鑫、黄森

展开 >

国家能源集团新疆开都河流域水电开发有限公司,新疆 库尔勒 841000

石河子大学水利建筑工程学院,新疆 石河子 832000

寒旱区生态水利工程兵团重点实验室,新疆 石河子 832000

集成经验模态分解 样本熵 长短期记忆网络 组合模型 日径流模拟

国家能源集团资助项目国家自然科学基金项目南疆重点产业创新发展支撑计划项目兵团科技创新人才计划项目

CEZB220403598521690052022DB0242023CB008-08

2024

10.13880/j.cnki.65-1174/n.2023.21.021

石河子大学学报(自然科学版)
石河子大学

石河子大学学报(自然科学版)

CSTPCD北大核心
影响因子：0.662
ISSN：1007-7383
年,卷(期)：2024.42(3)
  • 22