基于GPOS-BP神经网络模型的潮汐预报  

Tide Forecast Model based on GPOS-BP Neural Network

作　　者：秦思远 李进军 龙冰心 张正阳 王立泽 颜永豪 孙佳龙 QIN Siyuan;LI Jinjun;LONG Bingxin;ZHANG Zhengyang;WANG Lize;YAN Yonghao;SUN Jialong(Huaihai Institute Of Techology,Lianyungang 222000,China)

机构地区：[1]江苏海洋大学测绘与海洋信息学院,连云港222000

出　　处：《海洋信息》2020年第2期1-5,共5页Marine Information

基　　金：国家自然科学基金(40974016);自然资源部海洋信息技术创新中心开放基金课题;连云港高新区重点研发计划项目(ZD201905);江苏省高校海洋科学技术优势学科建设项目;测绘工程国家一流本科专业建设点建设项目。

摘　　要：为了提高预报潮汐的精度,提出了一种遗传算法(GA)与粒子群算法(PSO)结合的神经网络仿真模型进行潮汐预测,及引入遗传算法变异操作提升粒子群算法的寻优的性能,改善BP神经网络阈值和权值选取敏感、易陷入局部极小值的缺点。然后利用本文算法与调和分析法对潮汐进行非天文潮预报和天文潮预报。最后对Alexandria港口的实测潮汐数据仿真预测,实验结果验证了该方法的可行性与可靠性,且具有较高的预报精度。In order to improve the accuracy of tide forecasting,a genetic algorithm(GA)and particle swarm optimization(PSO)algorithm are combined with neural network simulation model.Genetic algorithm mutation operation is introduced to improve the performance of particle swarm optimization algorithm and improve BP neural network which has the shortage of threshold and weight selection are sensitive and easy to fall into the local minimum.Then the algorithm and harmonic analysis method are used for non-astronomical and astronomical tide forecast.Finally,the simulated tide data of Alexandria Port is verified by in-situ tidal observations.The experimental results confirm the feasibility and reliability of the method and its high forecasting accuracy.

关 键 词：潮汐预报 调和分析法 BP神经网络 粒子群算法 遗传算法 

分 类 号：P731.34[天文地球—海洋科学]