摘要
为提高叶尔羌河中长期径流预测精度,基于小波分析的基础上建立遗传算法优化BP神经网络的耦合模型,对60年叶尔羌河年径流时间序列进行研究。结果表明:耦合模型综合了两者的优势,在保留神经网络优良非线性拟合能力的同时,又融入遗传算法的容错性和全局搜索能力,提高预测径流时的学习速度和泛化能力。在对年径流进行预测时,其预测平均误差为-2.69%,而采用传统单纯的BP神经网络模型预测的平均误差为-10.25%。从预测误差检验以及模型的对比结果可知此模型合理、可行,因此该算法有助于解决叶尔羌河中长期径流预测问题。
In order to improve the prediction accuracy of middle and long term runoff in the Yarkant River,based on wavelet analysis,the coupling model of genetic algorithm to optimize BP neural network was established,and the annual runoff time series of 60-year Yarkant River was studied.The results showed that the coupling model integrated the advantages of the two methods,while retaining the good nonlinear fitting ability of neural network,it integrated the fault-tolerance and global searching ability of genetic algorithm,and improved the learning speed and generalization ability of predicting runoff.When the annual runoff was predicted,the average prediction error was-2.69%,while the traditional single BP neural network model was-10.25%.It can be seen from the prediction error test and the comparison results of the model that this model was reasonable and feasible.Therefore,this algorithm was helpful to solve the middle and long term runoff prediction problem of the Yarkant River.
作者
何兵
高凡
蓝利
覃姗
HE Bing;GAO Fan;LAN Li(College of Water Conservancy and Civil Engineering,Xinjiang Agricultural University,Urumqi,Xinjiang 830000)
出处
《安徽农业科学》
CAS
2019年第3期208-211,共4页
Journal of Anhui Agricultural Sciences
基金
新疆维吾尔自治区自然科学基金项目(2017D01A43)
关键词
径流预测
遗传算法
BP神经网络
叶尔羌河
Runoff prediction
Genetic algorithm
BP neural network
Yarkant River
