摘要
为了提高径流序列的稳定度和精度,减小参数优化不当导致的非线性误差,研究将长短期记忆神经网络(LSTM)、集成经验模态分解(EEMD)和北方苍鹰优化算法(NGO)相结合,构建了EEMD-NGO-LSTM耦合预测模型。将此预测模型应用于模拟东辽河中下游的控制总站——王奔水文站2012年~2021年逐月径流过程,并与鲸鱼算法(WOA)以及灰狼算法(GWO)优化的长短期记忆神经网络进行模型比较。结果表明,EEMD-NGO-LSTM耦合预测模型的超参数迭代速度最快,精度最高,预测结果最接近实测值,其决定系数R^(2)为0.8643。而后采用CMIP6气候模式(SSP126情景)下的2030年的降水、气温数据输入模型进行预测,在气温上升1℃,降水不变的情景下,年径流量将增加6.61%;在降水升高5%,气温不变的情景下,年径流量将增加6.95%;在气温上升1℃、降水升高5%的情境下,年径流量将增加22.16%。
In order to improve the stability and accuracy of runoff series and reduce the nonlinear error caused by improper parameter optimization,the long-and short-term memory neural network(LSTM),integrated empirical modal decomposition(EEMD)and Northern Goshawk Optimization Algorithm(NGO)are combined to construct a coupled EEMD-NGO-LSTM prediction model.This prediction model is applied to simulate the month-by-month runoff process from 2012 to 2021 at Wangbun Hydrological Station,the control terminus in the middle and lower reaches of Dongliao River,and the simulation results are compared with that of long-and short-term memory neural networks optimized by the Whale Optimization Algorithm(WOA)as well as the Gray Wolf Algorithm(GWO),respectively.The results show that the coupled EEMD-NGO-LSTM prediction model has the fastest hyperparameter iteration speed,the highest accuracy,and the closest prediction to the measured value,with a coefficient of determination R^(2)of 0.8643.The precipitation and temperature data for 2030 under the CMIP6 climate model(SSP126 scenario)are used to input into the model for the prediction,and the results show that,(a)under the scenario in which the temperature is increased by 1℃and the precipitation remains unchanged,the annual runoff will increase by 6.61%;(b)under the scenario with a 5%increase in precipitation and no change in temperature,annual runoff will increase by 6.95%;and(c)under the scenario with a 1℃increase in temperature and a 5%increase in precipitation,annual runoff will increase by 22.16%.
作者
张冲
王千凤
齐新虎
王思宇
陈末
ZHANG Chong;WANG Qianfeng;QI Xinhu;WANG Siyu;CHEN Mo(College of Water Resources and Electricity,Heilongjiang University,Harbin 150080,Heilongjiang,China;Cold Zone Groundwater Research Institute,Heilongjiang University,Harbin 150080,Heilongjiang,China)
出处
《水力发电》
CAS
2024年第1期1-7,共7页
Water Power
基金
黑龙江省普通本科高等学校青年创新人才培养计划(UNPYSCT-2020012)。
关键词
月径流预测
集成经验模态分解
北方苍鹰优化算法
长短期记忆神经网络
耦合模型
预测精度
monthly runoff prediction
integrated empirical modal decomposition(EEMD)
Northern Goshawk Optimization(NGO)
long-and short-term memory neural network
coupled model
prediction accuracy
