李天宇,吴远远,郝晓地,Meijer S. C. F.,翟学棚,刘杰,林甲.数据清洗对污水处理厂生物建模可靠性影响研究[J].环境科学学报,2020,40(9):3298-3310
数据清洗对污水处理厂生物建模可靠性影响研究
- Effect of data cleaning on the reliability of biologically modeling wastewater treatment processes
- 基金项目：国家自然科学基金项目（No.51878022）；国家水体污染控制与治理科技重大专项（No.2017ZX07102-003）
- 李天宇
- 北京首创股份有限公司, 技术中心/中-荷未来污水技术研发中心, 北京 100044
- 吴远远
- 北京首创股份有限公司, 技术中心/中-荷未来污水技术研发中心, 北京 100044
- 郝晓地
- 北京建筑大学, 北京未来城市设计高精尖中心/中荷未来污水技术研发中心, 北京 100044
- Meijer S. C. F.
- ASM Design B. V. /中-荷未来污水技术研发中心, 荷兰 3572 KX
- 翟学棚
- 北京首创股份有限公司, 技术中心/中-荷未来污水技术研发中心, 北京 100044
- 刘杰
- 北京首创股份有限公司, 技术中心/中-荷未来污水技术研发中心, 北京 100044
- 林甲
- 北京首创股份有限公司, 技术中心/中-荷未来污水技术研发中心, 北京 100044
- 摘要：为应对污水处理厂出水水质要求的不断提升，通过生物建模技术降低污水处理厂碳源、除磷药剂投加量，实现达标排放同时降低运营成本已成为必要手段.然而，数据质量是影响模型可靠性的关键因素.为提高建模数据质量，本研究在生物建模过程中提出了系统性3步清洗方法：统计分析法标定可疑数据；物料平衡计算法剔除大误差并闭合流量平衡；污泥特征分析法修正污泥组分.应用该方法成功对天津宁河芦台桥北污水处理厂2019年1—7月历史水质数据、SCADA在线流量数据及补充采集的3日水质数据实现有效的清洗，并完成基于Biowin-ASDM活性污泥模型的生物建模案例研究.通过与直接利用原始数据进行模拟拟合的对比结果表明，应用清洗后的数据进行建模可有效提升模型可靠性，不但校正参数趋于合理，更可实现生化段主要污染物TSS、TCOD、TN、TP去除率的模拟误差≤3%.
- Abstract：Due to the increasingly upgrading requirements on the effluent quality of wastewater treatment plants (WWTPs), the technique on biological modeling has gradually become an essential tool to reduce dosing both carbon for denitrification and chemicals for P-removal, due to the strict discharge standards and also expected low-operational costs. In this aspect, the quality of collected data is a key factor affecting the reliability of biological modeling. With this study, a three-steps method for data cleaning is systematically proposed to improve collected data for modeling:① highlighting suspicious data by statistical analysis; ② eliminating gross errors and closing flow balance by mass balance;③ correcting sludge components by sludge characteristics analysis. The proposed method has been successfully applied by the Biowin-ASDM model, in a case modeling study at the Lutaiqiaobei WWTP, Ninghe, Tianjin, with the collected data during January to July 2019, the SCADA on-line flow data and the additionally collected three-days data. Compared to the original data, modeling with the cleaned data can definitely improve the reliability of the biological model, which makes calibrated parameters reasonable but also the simulated error less than 3% on modeling the biological removal of TSS, TCOD, TN and TP.
