引用本文:
【打印本页】   【下载PDF全文】   【View/Add Comment】  【Download reader】  【 Close 】
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被：浏览 2537次   下载 1921次 本文二维码信息
码上扫一扫！
分享到： 微信 更多
字体:加大+|默认|缩小-
基于BP神经网络方法的近岸数值海温预报释用技术
匡晓迪, 王兆毅, 张苗茵, 何恩业, 邓小花
国家海洋环境预报中心 北京 100086
摘要:
为了提高近岸精细化海温预报精度，利用神经网络方法，分析了海温数值预报及观测数据在释用中的作用，研究了定点近岸海温影响因子的最优配置方案，建立了定点海温精细化数值预报释用模型，评估了释用模型性能。误差分析结果显示，数值海温产品及其观测在建模中起到了降低和稳定模型误差的作用；释用模型将定点数值预报的误差从2.2℃减少至0.7℃；预报误差较调训误差略高，但考虑到预报误差的稳定性，数值释用与人工经验预报水平持平，因此，该方法具有十分广阔的拓展空间和应用前景。
关键词:  数值预报  释用  人工神经网络  近岸
DOI：10.11693/hyhz20160700156
分类号:
基金项目:国家自然科学基金项目，41222038号；海洋公益性行业科研专项经费项目，201305031号。
AN INTERPRETATION SCHEME OF NUMERICAL NEAR-SHORE SEA-WATER TEMPERATURE FORECAST BASED ON BPNN
KUANG Xiao-Di, WANG Zhao-Yi, ZHANG Miao-Yin, HE En-Ye, DENG Xiao-Hua
National Marine Environmental Forecasting Center, SOA, Beijing 100086, China
Abstract:
To enhance the forecast accuracy in near-shore regions,we analyzed the effect of numerical forecast and observation on temperature,discussed the best configuration,established an numerical forecast interpretation model,and assessed the performance of the scheme.Error analysis shows that,the numerical products and observations are essential in reducing and stabilizing the error of the interpretation model.The interpretation model reduced the forecast error (2.2℃) to 0.7℃.Althouth slightly higher than the error of training model,the forecast errors are stable.Thus,the numerical interpretation could perform as empirical forecast,and may have great potential of application for near-shore sea forecast.
Key words:  numerical forecast  interpretation  artificial neural network  near-shore
Copyright ©  Editorial Office for Oceanologia et Limnologia Sinica    Copyright©2008 All Rights Reserved
Supervised by: China Association for Science and Technology   Sponsored by: Chinese Society for Oceanology and Limnology, Institute of Oceanology and Limnology, CAS.
Address: 7 Nanhai Road, Qingdao, China.    Postcode: 266071    Tel: 0532-82898753  E-mail: liuxiujuan@qdio.ac.cn  
Technical support: Beijing E-Tiller Co.,Ltd.