Abstract:

A novel cycle running activated sludge process named the five-tank integrated activated sludge process (FTIASP) was developed by Southeast University of China. It is similar to Unitank but has five tanks. Mathematical simulation and calibration of the FTIASP were performed, especially a iterative calculation procedure for calibrating kinetic parameters was proposed. The procedure combined six batch experiment simulations: ammonia uptake rate (AUR), nitrate uptake rate (NUR), oxygen uptake rate (OUR), phosphorus release rate (PRR), phosphorus uptake rate under aerobic (PUR_aerobic) and phosphorus uptake rate under anoxic (PUR_anoxic), sensitivity analysis and mathematical optimization method (genetic algorithm). The six batch experiments could simulate the nitrification, denitrification, aerobic growth of heterotrophic biomass, hydrolysis, phosphorus release and phosphorus uptake processes in FTIASP. Sensitivity analysis could calculate the influences of parameters on the model outputs, thus parameters needed to be calibrated were determined (parameter estimation). The mathematical optimization method could calibrate these parameters automatically. Good consistencies were found between the data of model prediction and experimental results, and the average relative deviation (ARD) values of the six batch experiments were 5.65%, 17.27%, 6.02%, 7.11%, 13.07% and 6.98% respectively, and those of effluent COD, NH₄⁺-N, TN and TP during dynamic simulations were 4.72%, 18.87%, 9.45% and 38.11%, respectively. The results indicated that the method proposed in this paper was feasible for calibrating the activated sludge model.

Key words: five-tank integrated activated sludge process, mathematical simulation, batch experiments, sensitivity analysis, genetic algorithm, parameter estimation

摘要：

对五箱一体化活性污泥工艺进行了数学模拟研究, 针对模型中的动力学参数提出了一种迭代计算的校正方法。该方法结合了6个批式实验模拟(氨氮吸收速率AUR、硝氮吸收速率NUR、耗氧速率OUR、厌氧释磷速率PRR、好氧吸磷速率PUR_aerobic和缺氧吸磷速率PUR_anoxic)、灵敏度分析以及数学优化方法(遗传算法):通过迭代计算确定合适的污泥组分比例, 完成对6个批式实验的模拟;灵敏度分析可以分别确定各个批式实验模拟中参数对模拟结果的影响程度, 挑选待优化参数, 完成参数识别;数学优化方法可以自动对参数进行校正。结果显示, 校正后的模型对五箱一体化活性污泥工艺的模拟效果较好:6个批式实验的模拟值与测量值之间的平均相对误差(ARD)分别为5.65%、17.27%、6.02%、7.11%、13.07%和6.98%;30 d动态出水COD、NH₄⁺-N、TN和TP的ARD值分别为4.72%、18.87%、9.45%和38.11%。研究结果表明, 提出的迭代方法可以用来对活性污泥模型进行校正, 而且校正效果较好。

关键词: 五箱一体化活性污泥工艺, 数学模拟, 批式实验, 灵敏度分析, 遗传算法, 参数识别