Soft-Sensing, Automation, and Diagnosis for Nitrification

Abstract

Nitrification of wastewater requires control for the purposes of efficiency and of stability. However, first, the highly complex microbial community in wastewater treatment plants and very different compositions of wastewaters impede a complete understanding of nitrification. Second, the determination of the states of the process is seldom exact because sensors are subject to a high level of wear and tear. Consequently, instrumentation, control, and automation (ICA) of nitrification processes are a challenge. I hypothesize that soft-sensors - based on the signals of robust, physical sensors - and specially designed control loops, which do not need redundant information to correct for drift, can reduce the maintenance needs for the control of nitrification. Consequently, the econonomical application range for ICA of nitrification processes can be extended. A large part of this thesis is dedicated to the stabilisation of urine nitrification as a pretreatment for nutrient recovery. To this end, an online nitrite detection and an according control of the process are required. Furthermore, the thesis comprises an ammonia aeration controller, which is based on pH signals, to be used in municipal wastewater. The four chapters of the thesis focus on the following topics. The first chapter investigates calibration experiments for in-line UV-Vis sensors. Such sensors can serve as soft-sensors for numerous substances. With differently designed experi- ments, data for the calibration of UV-Vis based nitrite models in high strength wastewater was recorded. Online experiments executed in-situ allow calibrating similarly exact models like experiments which are conducted by spiking additional nitrite. The good performance of the models being based on online experiments is explained by the correlation of other light-absorbing substances in the process with nitrite. The second chapter evaluates the potential of a stoichiometry-based soft-sensor for nitrite. The stoichiometry of nitrification theoretically allows distinguishing the ammonia and the nitrite oxidation rates by means of the derivatives of the dissolved oxygen and of the pH signal and in turn detecting nitrite accumulations. We show that under relatively ideal conditions the dynamics in the ratio of the two derivatives show similarities with the dynamics in the nitrite concentration. Factors that deteriorate this information quality are identified. The third part of this thesis investigates how drifting sensor signals can be used as an input for a feedback controller without relying on redundant information to correct this drift. The controller shall prevent uncontrolled nitrite accumulation in a urine nitrification system. To this end, the controller does not use the absolute value of the nitrite signal, but its 1st and 2nd derivative as inputs. By controlled accumulation and degradation of nitrite, the controller can keep the derivatives continuously informative. The switch from substrate inhibition to substrate limitation of the nitrite oxidising bacteria can be identified by means of the inflection point in the signal. The developed controller is tested successfully on a lab-scale reactor. The fourth chapter of this thesis describes an ammonia based aeration controller. These controllers usually require maintenance intensive instrumentation. Therefore a new am- monia soft sensor for continuously operated municipal wastewater treatment plants is developed. It relies on the 1st derivative of a pH difference measured over the aerated zone. This soft sensor is successfully tested in a controller deployed on various municipal wastewater treatment plants. The controller was designed to increase its robustness towards drift in the pH signals.