人工神经网络在岛屿近岸海浪模拟中的应用

Application of artificial neural network to numerical wave simulation in the coastal region of island

以三亚东锣岛近岸海域为研究案例,基于9个月现场观测数据,利用人工神经网络模型对区域海浪模式WAVEWATCHⅢ(WW3)的有效波高结果进行了修正.结果显示:受空间分辨率的限制,WW3模式难以有效地刻画岛屿岸线和地形对波浪的影响,导致有效波高的模拟结果与现场观测存在较大差距;而使用该文建立的人工神经网络模型则能有效地修正有效波高的模拟结果,使其与观测值的均方根误差和平均绝对误差明显下降.进一步分析计算显示,当人工神经网络模型应用于岛屿近岸波浪要素的模拟修正时,有必要将平均波向作为输入变量,使不同来向的波浪传至岛屿近岸所经历过程的差异性得以体现.该方法及相关结果对岛屿近岸长时间序列波浪要素信息的模拟和预报具有工程应用价值.

Using the coastal region of Dongluo Island,Sanya as a case study,the significant wave height of the regional wave model WAVEWATCHⅢ(WW3)is modified by using the ANN model based on 9-month observational data.Results show that due to the limitation of spatial resolution,it is difficult for the WW3 model to effectively describe the influence of island shoreline and terrain on waves,leading to a large gap between modeling results and field observations for the significant wave height.The artificial neural network(ANN)model established in this paper can effectively correct the modeling results,with the root mean square error and mean absolute error obviously reduced.The calculations further show that when ANN model is applied to the wave elements modeling in the coastal region of island,it is necessary to take the average wave direction as the input variable to reflect the disparity of waves propagating processes from different directions to the coastal region of island.The method and the results are helpful for the simulation and prediction of long-term wave elements in engineering application for the coastal region of island.