Wang Aijie, Liu Chunshuang, Han Hongjun, Ren Nanqi, Lee Duu-Jong
State Key Laboratory of Urban Water Resource and Environment (HIT), Harbin, China.
J Hazard Mater. 2009 Sep 15;168(2-3):1274-9. doi: 10.1016/j.jhazmat.2009.03.006. Epub 2009 Mar 14.
The denitrifying sulfide removal (DSR) process has complex interactions between autotrophic and heterotrophic denitrifers; thus, constructing a detailed mechanistic model and proper control architecture is difficult. Artificial neural networks (ANNs) are capable of inferring the complex relationships between input and output process variables without a detailed characterization of the mechanisms governing the process. This work presents a novel ANN that accurately predicts the steady-state performance of an expended granular sludge bed (EGSB)-DSR bioreactor for nitrite denitrification and the complete DSR process. The proposed ANN shows that at a threshold hydraulic retention time (HRT)<7h, influent sulfide concentration markedly affects reactor performance.
反硝化除硫（DSR）过程中自养型和异养型反硝化菌之间存在复杂的相互作用；因此，构建详细的机理模型和合适的控制架构具有一定难度。人工神经网络（ANN）能够推断输入和输出过程变量之间的复杂关系，而无需对控制该过程的机制进行详细描述。这项工作提出了一种新型人工神经网络，它能够准确预测用于亚硝酸盐反硝化和完整DSR过程的膨胀颗粒污泥床（EGSB）-DSR生物反应器的稳态性能。所提出的人工神经网络表明，在水力停留时间（HRT）阈值<7小时时，进水硫化物浓度会显著影响反应器性能。
