基于数据驱动模型的潮位和潮流预测方法研究

Tidal Level and Current Prediction on the Basis of Data-Driven Model

为解决工程海域潮位、潮流资料不足给海洋工程设计和数学模型建立带来的不确定性,根据单测站潮位潮流的自相关性、对应测站潮位或潮流以及潮位与潮流之间的互相关性,建立基于数据驱动模型人工神经网络的单测站潮位、潮流(流速、流向)预测模型;多测站潮位、潮流对应预测模型;潮位与潮流对应预测模型.以复杂海况下的实测潮位、潮流资料进行模型验证,重现了潮位、潮流自身及相互之间的非线性映射关系.模型预测结果与现场实测数据的比较及其误差分析表明,该模型具有结构简单、精度高的优点,适用于解决工程实际问题.

The insufficiency of tidal level and current data near the ocean engineering waters may bring uncertainty for ocean engineering design and numerical model construction.To solve this problem,some of necessary prediction models are developed,including one site tidal level or current（current velocity and current direction） prediction model,multi-site tidal level or current prediction model,and tidal level-current prediction model.The construction of those models is based on the back propagation（BP） artificial neural network（ANN） and the properties of self-correlation of single site and cross-correlation among multi-sites or between tidal level and current.Field data under complex geography and hydrodynamic condition are used to validate the performance of the presented data-driven models.It is indicated that the nonlinear mapping relation between tidal level and tidal current can be reproduced by those models.The comparison between the numerical results and field-data verifies that the developed models have the advantages of simple structure and good precision.