摘要
为进一步改善长期多因子径流模拟效果,提出了基于马尔科夫链修正的PSO和Elman耦合的径流模拟模型。该模型采用PSO算法优化Elman模型的参数,然后将优化后的参数值分配给Elman模型作为网络训练的参数,再运用马尔科夫链对初始预测值进行修正,得到最终预测值。将提出的模型应用于松花江支流呼兰河兰西水文站的年径流深模拟预测中,并与传统Elman模型、简单线性回归模型、PSO-Elman模型进行对比。结果表明:优化参数后的模型预测效果优于传统神经网络模型和简单线性回归模型,PSO-Elman模型较传统Elman模型平均相对误差和均方根误差减少了49.1%,30.2%,确定性系数由0.32提升至0.67;较简单线性回归模型平均相对误差和均方根误差减少了61.2%,37.7%,确定性系数由0.14提升至0.67;经马尔科夫链模型残差修正后,PSO-Elman组合预测模型平均相对误差和均方根误差减少了57.7%,52.2%,确定性系数由0.67提升至0.92。提出的模型不但精度有显著提高,且计算过程简便,是一种有较强应用价值的年径流预报模型。
To further improve the long-term multi-factor runoff simulation effect,this paper proposes a PSO-Elman coupling runoff simulation model modified by Markov chain.Firstly,PSO algorithm is used to optimize the parameters of Elman model.Then,the optimized parameter values are assigned to Elman model as the initial parameters of network training.Finally,Markov chain is used to modify the predicted value to obtain final predicted value.In this study,the proposed model is applied to the simulation and prediction of annual runoff depth of Lanxi hydrological station on Hulan River,a tributary of Songhua River,and contrasted with the traditional Elman model,the simple linear regression model and PSO-Elman model.The results show that the prediction effect of the optimized model is better than that of the traditional neural network model and the simple linear regression model.Compared with the traditional Elman model,the mean relative error and root mean square error of PSO-Elman model are reduced by 49.1%and 30.2%,and the deterministic coefficient is increased from 0.32 to 0.67.Compared with the simple linear regression model,the mean relative error and root mean square error are decreased by 61.2%and 37.7%,and the deterministic coefficient is increased from 0.14 to 0.67.After the residual correction of Markov chain model,the mean relative error and root mean square error of PSO-Elman combined prediction model are reduced by 57.7%and 52.2%,and the deterministic coefficient is increased from 0.67 to 0.92.Therefore,the proposed model not only has significantly improved accuracy,but also has a simple calculation process.It is a kind of annual runoff prediction model with strong application value.
作者
王文川
王莉芳
郭安强
WANG Wenchuan;WANG Lifang;GUO Anqiang(College of Water Resources,North China University of Water Resources and Electric Power,Zhengzhou 450046,China;Henan Hebi Hydrology and Water Resources Survey Bureau,Hebi 458030,China)
出处
《人民长江》
北大核心
2022年第11期66-71,共6页
Yangtze River
基金
河南省重点研发与推广专项(202102310259,202102310588)
国家自然科学基金项目(51509088)。
