污水生物处理建模水质数据误差来源分析与影响评价

为进一步提高国内基础建模数据质量，降低建模实践中的数据清洗难度，本研究以北京龙庆再生水厂生物建模过程中的实际数据问题为例，基于数理统计理论及国内污水水质特征分别构建误差来源分析方法与误差影响评价方法.通过设计补充采样与实验排查方案，成功识别数据误差来源并实现不同类型误差对数据准确性的影响量化评价.研究结果表明，污水处理厂日常的取样与化验过程中产生的随机误差不可忽视，特别是当大量随机误差存在时，对数据总体造成的波动影响无法通过数据量的积累而消除；而不同的采样位置与不当的样品保存方式则会引起水质发生变化，由此产生的系统误差将导致测量结果长期偏高/偏低.因此，在依赖人工测量数据进行建模的国内实践中，应首先对历史数据进行误差检验，再进行必要的建模补充采样，以避免补充数据质量受相同误差影响，减少非必要的人力、物力浪费.

Error source detection and effect evaluation of water quality data for modeling biological wastewater treatment

To improve the original data quality of modeling biological wastewater treatment and to reduce the difficulty of data cleaning in modeling, the study applied a practical data (from Beijing Longqing wastewater treatment plant) problem as an example in modeling, and established both error source detection and error effect evaluation based on the mathematical statistical theory and the domestic wastewater characteristics. With the help of the supplementary sampling and experimental investigation, the data errors from different sources were successfully identified, and the effects on the data accuracy by different error were quantitatively analyzed. The results indicate that random errors in the daily sampling and lab testing process cannot be neglected, and the effect of fluctuations caused by the whole data cannot be eliminated by the data accumulation, particularly when a large number of random errors exist. However, different sampling locations and improper samples'' storage methods would cause some changes of water quality, which could result in a systematic error that would further lead to long-term high/low measurement results. Therefore, the historical data should be first checked for their errors, and then necessary supplementary sampling is carried out when manual measurement is used for model application, to avoid the quality of supplementary sampling data affected by the same errors and to reduce unnecessary labors and materials.