遗传门限自回归模型在感潮河段水位预测中的应用

Application of genetic threshold auto-regressive model to water stage forecasting for tidal river section

应用门限自回归(TAR)模型建立了同时受潮汐和径流双重影响的长江下游感潮河段高桥水文站月水位TAR预测模型,建模过程中运用遗传算法来实现模型参数的优化.计算结果显示,门限自回归模型可以拟合感潮河段的非线性特性,拟合及预测精度均满足水文预报规范要求,遗传算法的引入简化了建模过程,提高了模型的预测精度并保证了其预测性能的稳定性.研究结果表明用遗传门限自回归模型预测感潮河段的水位是可行的,该模型在感潮河段其他水文要素的非线性时序预测中也具有广泛的实用价值.

A threshold auto-regressive （TAR） model was established for forecasting monthly water stage at the Gaoqiao hydrological station in the lower tidal reaches of the Yangtze River, which is under the impacts of both tides and runoff. The genetic algorithm was adopted for parameter optimization in modeling of TAR. The calculated result shows that the TAR model can effectively reflect the nonlinear characteristics of tidal rivers, and the precision of fitting and simulation accords with the standard of hydrological forecasting. The application of the genetic algorithm simplifies the process of modeling, improves the forecasting accuracy, and ensures the high stability of the model. Therefore, the TAR model based on the genetic algorithm for water stage forecasting of tidal rivers is practical and has high adaptability to forecasting of other hydrological nonlinear time series in tidal rivers.