城市排水管网外水诊断中特征因子的选择

为减少城市排水管网外水诊断中选择特征因子时对人为经验的依赖性,需建立特征因子选择的判别准则。解释特征因子需满足显著性差异、独立性、较小方差等要求的本质原因。利用Mann-WhitneyU检验、Pearson相关性检验等统计方法,给出特征因子显著性差异和独立性判断的定量标准,以及通过水质数据标准差与误差之间的关系确定特征因子标准差上限的方法,并提供相应的例子。结果表明,当显著性差异水平和相关系数水平小于0.05的水质指标才能作为特征因子,同时当水质样本为30~60个,相对误差控制在5%~10%时,特征因子的变异系数应小于0.3。

In order to reduce the human experience dependency in the selection of marker species in the diagnosis of extraneous water of urban drainage network,it is necessary to establish a criterion for the selection of marker species.The reasons why the marker species need to meet the requirements of significant difference,independence and small variance are explained.By using statistical methods such as Mann-WhitneyU test and Pearson correlation test,the quantitative criteria for judging the significance difference and independence of characteristic factors are given,and the method of determining the upper limit of standard deviation of characteristic factors through the relationship between standard deviation and error of water quality data is given,and corresponding examples are provided.The results show that only the water quality index with significant difference level and correlation coefficient level less than 0.05 can be used as characteristic factor,and when the number of water quality samples is 30~60 and the relative error is controlled at 5%-10%,the coefficient of variation of characteristic factor should be less than 0.3.