自回归模型在枯季径流模拟预测中的应用

Application of Autoregressive Model for Low Flow Prediction

摘要: 枯季水文径流预测是现在水文预测预报的一个重要组成部分，随着社会经济发展和人口增加，水资源问题越来越突出，因此开展枯季水文径流模拟预测为准确把握流域枯季水资源水量和水文过程提供了依据。西江流域是珠江水系的第一大支流，近年来由于径流量减少，特别是枯季径流变化较大，珠江三角洲河口咸潮上溯年年发生，影响区域生产生活，因此有必要开展枯季径流模拟预测研究。自回归模型是一种基于时间序列的预测预报方法，为了研究模型在此区域枯季径流模拟预测的适用性，采用自回归模型对贵港站日平均流量进行研究与分析，所率定的自回归阶数p，经验证是合适的，模拟预测结果表明：洪峰相对误差小于20%，径流深相对误差小于5%，确定性系数值大于0.75，精度较好。说明自回归模型在贵港水文站枯季径流模拟预测中是适用的。

Abstract: The water resources problem becomes increasingly prominent due to the development of socio- economic and population growth. Low flow prediction is becoming important since it can provide based evidence for the water resources quantity and hydrological processes in dry-season. The estuarine salt tide in the Pearl River Delta traced occurs every year due to the reduction of runoff, which affects the regional production and life. The autoregressive model is used to simulate and predict daily average flow at Guigang hydrologic station. The self-regression order parameter of p is obtained and tested. Results show that the relative difference of peak flow is less than 20%, the relative error of runoff depth is less than 5%, and the uncertainty coefficient value is greater than 0.75. This shows that the auto-regressive model is applicable for the low flow simulation and prediction in the Guigang hydrological station.