引用本文:
【打印本页】   【下载PDF全文】   查看/发表评论  下载PDF阅读器  关闭
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 1823次   下载 1325 本文二维码信息
码上扫一扫!
分享到: 微信 更多
黄河源区河川径流短期预测的ANFIS模型
马盼盼1, 白 涛1, 武连洲,等1
西安理工大学 西北旱区生态水利工程国家重点实验室培育基地
摘要:
【目的】建立径流短期预测的自适应神经模糊推理系统(ANFIS)模型,以提高预测精度,进而为黄河源区水资源开发和工程规划提供参考。【方法】以黄河源区出口站军功水文站为研究对象,以ANFIS为基本方法,建立ANFIS日尺度径流预测模型。基于输入变量、训练次数、隶属度函数类型与数目、预见期等参数设置了9个方案,通过实测径流与预测径流的对比和评价指标(均方根误差RMSE、相关性系数R)验证确定最佳方案,并分析不同参数对预测结果的敏感性,获得基于最优参数的ANFIS模型。【结果】采用神经网络+Sugeno型模糊推理算法建立了ANFIS日尺度径流预测模型,在预见期为1 d时,利用ANFIS模型进行的径流短期预测,其相对误差最大为4.36%,平均为0.21%,预测结果合理可靠;当预见期延长至2~4 d时,预测结果均满足精度要求,相对误差平均值均小于3.00%。【结论】将ANFIS用于短期径流预测,既可提高预测精度,又能延长预见期,可为黄河源区水库群规划、施工、调度和全流域水资源配置提供指导。
关键词:  自适应神经模糊推理系统  径流预测  敏感性分析  隶属度函数  黄河源区
DOI:
分类号:
基金项目:国家自然科学基金项目(51409210,91325201);水利部公益性行业科研专项(201501058);陕西省水利科技计划项目(2016slkj-8,2017slkj-16);西安理工大学水利水电学院科研项目(2016ZZKT-15)
Short-term runoff prediction by ANFIS model in source region of the Yellow River
MA Panpan,BAI Tao,WU Lianzhou,et al
Abstract:
【Objective】This study established an adaptive network-based fuzzy inference system (ANFIS) to improve the accuracy of runoff forecast and provide reference for water resources development and project planning in source region of the Yellow River.【Method】Jungong station,located at the outlet of source region of the Yellow River,was chosen as the target and daily runoff forecast model was established by ANFIS.Nine schemes were set up based on input variables,training times,types and numbers of membership functions and forecast period.This study recommended the best scheme through contrastive analysis of measured and predicted runoff as well as statistical indexes including RMSE and R.The ANFIS model with optimal parameters was also obtained by analyzing the sensitivity of forecast results to different parameters.【Result】A daily runoff forecast model was established by using neural network and Sugeon fuzzy inference algorithm.When the expected period was 1 day,the maximum error of the ANFIS model was 4.36%,the average relative error was 0.21%,and the results were acceptable.When forecast period of the model was extended to 2-4 days,the average relative error was less than 3.00% and the forecast results met the precision requirements.【Conclusion】Using ANFIS for short-term runoff forecast not only enhanced the prediction accuracy,but also extended the forecast period.It would provide guidance for cascades planning,reservoir construction,operation and water allocation in the whole Yellow River Basin.
Key words:  adaptive network-based fuzzy inference system  runoff predication  sensitive analysis  membership functions  source region of the Yellow River