Model based predictive control for energy efficient biological nitrification process with minimal nitrous oxide production

Recent studies reveal that Ammonium Oxidizing Bacteria (AOB) in the Biological Nitrification Removal (BNR) process is one of the main contributors for Nitrous Oxide (N2O) emissions, a powerful greenhouse gas having a potential of 300times that of Carbon Dioxide (CO2) (IPCC, 2011; Ravishankara et al., 2009 [1,2]). Though a few models have been proposed to understand the behaviour of N2O production by AOB under various conditions, there exists hardly any work that aim towards development of a control strategy that utilizes these kind of models to mitigate N2O production. In this work, a model is developed based on the experimental studies (Ni et al., 2013 [3]) that capture the dynamics of N2O during recovery to aerobic conditions, after a period of anoxia, a common practice in nitrogen removal process. Subsequently, this model is employed in soft sensing using Extended Kalman Filter (EKF) to estimate N2O concentration and develop an advanced model based control strategy for energy efficient BNR process with minimal N2O production. This control strategy provides an aeration profile that minimizes N2O production and energy consumption (less pumping cost) apart from meeting the desired ammonium level at the output. The key features of the proposed model based control strategy are: (i) only continuous measurements of DO is required and, (ii) fairly insensitive to fluctuations in the influent ammonium loading and changes in the model parameters.

近期研究表明，生物硝化去除（BNR）过程中的氨氧化细菌（AOB）是氧化亚氮（N₂O）排放的主要贡献者之一，氧化亚氮是一种强效温室气体，其潜在影响是二氧化碳（CO₂）的300倍（政府间气候变化专门委员会，2011年；拉维尚卡拉等人，2009年[1,2]）。尽管已经提出了一些模型来了解AOB在各种条件下产生N₂O的行为，但几乎没有任何工作旨在开发一种利用此类模型来减少N₂O产生的控制策略。在这项工作中，基于实验研究（倪等人，2013年[3]）开发了一个模型，该模型捕捉了在缺氧一段时间后恢复到好氧条件期间N₂O的动态变化，这是脱氮过程中的常见做法。随后，该模型在使用扩展卡尔曼滤波器（EKF）的软测量中被用于估计N₂O浓度，并为具有最小N₂O产生的节能BNR过程开发一种基于先进模型的控制策略。这种控制策略除了在输出端满足所需的氨水平外，还提供了一种曝气模式，可使N₂O产生和能源消耗（更低的泵送成本）最小化。所提出的基于模型的控制策略的关键特征是：（i）只需要连续测量溶解氧（DO），以及（ii）对进水氨负荷的波动和模型参数的变化相当不敏感。