Model Predictive Control Strategy for Optimizing Biological Nitrogen Removal (BNR) Processes Accounting for Greenhouse Gas Emissions

Biological Nitrogen Removal (BNR) process comprises sequential oxidation of ammonia to nitrate and subsequent reduction of nitrate to nitrogen gas under a sequence of aerobic and anoxic conditions. Ammonia oxidizing bacteria (AOB) which are used for nitrification are the main contributors of Nitrous Oxide (N2O), a powerful greenhouse gas having a potential of 300 times greater than Carbon Dioxide (CO2) [1] and Nitric Oxide (NO), which is a toxic gas. Due to unavailability of unified model for capturing the dynamics of N2O it is difficult to control its emission from waste water plants. In this study, a model is chosen that captures the dynamics of N2O during recovery to aerobic condition after a period of anoxia (which is a common practice in waste water treatment plant) that is used for control purposes. Further, many of the states (like cell concentration, nitrous oxide and nitric oxides) used in the model cannot be or are expensive to measure (unknown states) in a real BNR process. In order to mitigate the emission of N2O its concentration is first estimated with a soft sensor (Extended Kalman Filter) and then a nonlinear model predictive control is implemented. Finally, a control algorithm is provided to address a multi objective problem such as mitigation of liquid N2O ( 0:001(mg=L)), maintaining DO (2(mg=L)) and NH+

4 concentration (1(mg=L)) [2] in effluent water.