• 关键词：污水处理  进水特征  统计分析  Prophet时序模型  模拟预测

• 基金项目：国家水体污染控制与治理科技重大专项（No.2017ZX07102-003）

• 邱勇
• 清华大学环境学院，北京 100084

• 毕怀斌
• 清华大学环境学院，北京 100084

• 田宇心
• 清华大学环境学院，北京 100084

• 梁鹏
• 清华大学环境学院，北京 100084

• 黄霞
• 清华大学环境学院，北京 100084

• 摘要：污水处理厂进水水量和水质呈现复杂的变化模式，准确掌握进水数据的变化特征已经成为改善工艺运行的关键.已有定性分析方法一般难以准确解释和预测变化特征，而已有定量分析模型常常结构复杂和实现难度大，需要针对应用需求进行改进.本研究使用统一模型描述进水特征并简化了实现方式.基于经典统计分析和开源Prophet模型，进行了污水处理厂进水数据的预处理、统计、建模和可视化.最后使用 3座污水处理厂的实际数据，识别了多重季节性、长期趋势、节假日等的影响特征，说明了统一模型对污水处理厂进水特征的解释和预测能力.结果可供污水处理工艺优化运行和环境时序数据解析等参考.

• Abstract：The variation mode of influent flowrate and water quality to wastewater treatment plants （WWTPs） is usually too diverse to predict， thus good understanding of the influent characteristics （as the variation of water quality and quantity） has been the bottleneck to optimize the process operation. Currently qualitative analyses often failed to explain the WWTP influent characteristics in satisfied accuracy， whereas quantitative models might also fail in prediction due to their complexity in model structure and program coding. Therefore， for the sake of better process operation in actual facilities， it is necessary to improve the influent characterization models. Here， an integrated model was proposed， based on classical statistical analyses and open-sourced Prophet model， to simplify the analyses and visualization of the influent data. Using this model， we applied the data description， statistics， modeling and visualization in a stand-alone program， and explained the influent characteristics by multiple seasonal， long-term and holiday trends. Three case-studies were given to evaluate the performance of the integral model in predicting the influent flowrate and COD variations of the WWTPs. The results could support the process optimization in WWTP as well as the time-series data analysis for environmental issues.

• Key words：wastewater treatment  influent characteristics  statistical analysis  Prophet model  model-based prediction

• 
  摘要点击次数： 294 全文下载次数： 519