Parameter Identification of an Ultrafiltration Model for Organics Removal in a Full-Scale Wastewater Reclamation Plant with Sparse and Incomplete Monitoring Data.
Parameter Identification of an Ultrafiltration Model for Organics Removal in a Full-Scale Wastewater Reclamation Plant with Sparse and Incomplete Monitoring Data.
复制标题
监测数据稀疏且不完整的全规模废水回收厂去除有机物的超滤模型参数辨识
DOI：
10.1371/journal.pone.0161300
复制
发表时间：
2016
期刊：
影响因子：
3.7
通讯作者：
He M
中科院分区：
文献类型：
--
作者：
Sun F;Zeng S;Huang Y;He M
超滤技术已成为我国污水资源化的主要处理工艺之一。建模是理解和优化超滤系统的有效工具。为此，以前开发的有机物去除超滤模型应用于超滤过程中的一个典型的，全面的污水再生厂（WRP）在中国。然而，稀疏和不完整的现场监测数据从所研究的WRP使传统的模型分析方法在这种情况下很难工作。因此，提出了两种策略，即策略1和策略2，根据区域敏感性分析方法，用于模型参数识别。策略1旨在同时识别模型参数和缺失的模型输入，即采样时间，而策略2试图通过迭代过程将这两个过程分开以降低识别问题的维度。在这两种策略下，模型在清河水资源规划中表现良好，总有机碳（TOC）模拟值与实测值之间的绝对相对误差一般在10%以下。四个模型参数均具有较好的敏感性和可识别性，甚至可以粗略识别采样时间。在模型输入不完全的情况下，这些结果是令人鼓舞的，并增加了模型的可信度，当它被应用到清河WRP。
Ultrafiltration (UF) has become one of the dominant treatment processes for wastewater reclamation in China. Modeling is an effective instrument to understand and optimize UF systems. To this end, a previously developed UF model for organics removal was applied to the UF process in a typical, full-scale wastewater reclamation plant (WRP) in China. However, the sparse and incomplete field monitoring data from the studied WRP made the traditional model analysis approaches hardly work in this case. Therefore, two strategies, namely Strategy 1 and Strategy 2, were proposed, following a regional sensitivity analysis approach, for model parameter identification. Strategy 1 aimed to identify the model parameters and the missing model input, i.e. sampling times, simultaneously, while Strategy 2 tried to separate these two processes to reduce the dimension of the identification problem through an iteration procedure. With these two strategies, the model performed well in the Qinghe WRP with the absolute relative errors between the simulated and observed total organic carbon (TOC) generally below 10%. The four model parameters were all sensitive and identifiable, and even the sampling times could be roughly identified. Given the incomplete model input, these results were encouraging and added to the trustworthiness of model when it was applied to the Qinghe WRP.