确定河渠纵向离散系数的分位数回归法

Quantile regression method to determine longitudinal dispersion coefficient of river channel

采用分位数回归的纵向离散系数研究方法和双站点浓度时间数据,对突发水污染事故中河渠的水质进行预测,并对比分析了分位数回归与最小二乘法回归效果。实例研究结果显示,运用分位数回归法确定河渠纵向离散系数效果好,第一站点的回归参数通过了97.5%置信水平下的假设检验,第二站点的预测值与实际值相关系数最高达到了0.928。同时,分位数回归法在解决偏态分布问题时较最小二乘法有明显优势。

The determination method of longitudinal dispersion coefficient based on the quantile regression and the concentration-time data of two sites were used to predict the water quality in the river channel. Moreover, the results determined by the quantile regression and least squares regression methods were compared. The example results showed that the quantile regression method is feasible and effective to determine the longitudinal dispersion coefficient of river channel. The quantile regression parameters passed the hypothesis test in the 97.5% confidence level at the first site, and the highest correlation coefficient of the predicted and actual values at the second site reached up to 0.928. In addition, quantile regression method has more advantage in solving the problems with skewed distribution than the least squares regression method.