【摘要】 利用2006—2017年我国南海部分区域(112°～114°E,10°～12°N)的Argo观测数据,对该海区声速剖面进行了仿真分析和研究。在此基础上,利用遗传算法(GA)优化的径向基函数(RBF)神经网络建立反演预测模型(GA-RBF),结合海区表面实测温度和历史数据,研究了该区域2016—2017年的声速剖面实时预测情况,并获得该海区6月和12月的声速剖面平均均方根误差值为0.845 m/s和0.815 m/s;而采用平均声速剖面方法获得该海区6月和12月的声速均方根误差分别是2.393 m/s和2.176 m/s。仿真结果表明:基于GA-RBF网络模型并利用海区表面实测温度的反演预测结果更趋近实测声速剖面,该模型可用于海区垂直声速剖面的实时预测。更多还原

【Abstract】 Based on Argo data in a region(112°～114°E, 10°～12°N) of the South China Sea from 2006 to 2017, the ocean SSP is analyzed to obtain Empirical Orthogonal Function(EOF). The inversion prediction model for ocean SSP is developed with Radial Basis Function(RBF) neural network optimized by Genetic Algorithm(GA)(GA-RBF). This model is used to inverted SSP in 2016—2017 with Sea Surface Temperature(SST). The averaged Root-Mean-Square Errors(RMSE) of inverted SSP in June and December are 0.845 m/s and 0.815 m/s with GA-RBF, and those are 2.393 m/s and 2.176 m/s with AGV. The GA-RBF model is better and it can be used to real-time forecast the SSP.更多还原