珠江口磨刀门水道盐度多步预测研究

Salinity Prediction at Modaomen Waterway in Estuary of Pearl River

为了减少咸潮灾害对人民生产生活的影响,进行盐度多步预测研究对建立有效的咸潮预警机制具有十分重要的意义。本文利用引入滞后因子的DE-RBF方法,以珠江口磨刀门水道为例,研究建立盐度多步预测模型的可行性。研究表明:滞后因子的引入使得预测模型的精度进一步提高,采用的DE搜索策略可以快速寻找到全局最优值。建立的5步盐度预测模型在100mg/L的允许误差范围内拟合度达到68.91%,而在200mg/L的情况下拟合度达到80.79%。而且这种方法具有普适性,在靠近外海的挂定角站可得到相似的结论。联石湾站和挂定角站的5步预测模型的预警准确度均超过90%,而预警错误率分别为9.47%和27.03%。

The research on salinity prediction at Modaomen waterway in the estuary of the Pearl River was con-ducted by building multistep salinity forecasting model via RBFNN.Lag operator and differential evolution algo-rithm were brought into search and build the optimal one-step forecasting model,and then multistep forecasting model was built.The results show that the salinity forecasting model is more precise with the introduction of lag operator,and DE can find the optimal value in just a few iterations.The 5-step salinity forecasting model can reach its goodness of fit high up to 68.91% within the permissible error range of 100mg/L,while 80.79% if the permissible error bound is set to 200mg/L.The similar conclusion can also be reached when the method above is applied to salinity series at Guadingjiao Station in the lower reaches near the open seas.Precautionary accuracy of 5-step forecasting model can reach up to 90% at the both stations of Lianshiwan and Guadingjiao,while the error rate is only 9.47% and 27.03%.