Salinity Prediction at Modaomen Waterway in Estuary of Pearl River
Chen Zishen · Journal of China Hydrology · 2010
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%.