Masters Theses

Date of Award

5-1997

Degree Type

Thesis

Degree Name

Master of Science

Major

Chemical Engineering

Major Professor

Tse-Wei Wang

Committee Members

Charles Moore, Chris Cox

Abstract

The focus of this study is to evaluate and compare two alternate control strategies for a multi-stage wastewater treatment plant, in the attempt to offer insight into what benefits first level automated closed-loop control may have over manual open-loop control. The plant under study is being built in Victoria, Texas by the DuPont Corporation, with the potential of several other DuPont plants being modeled after this one. The evaluation and comparison of the alternate control strategies is based upon the ability of each control system to render the overall closed-loop system stable over long term, and to meet plant treatment performance criteria in the face of setpoint and disturbance perturbations. The ability to maintain stability may come as a sacrifice of some degree of desirable performance. The greater the margin of stability, the poorer the performance, and vice versa. The ease of controller tuning is also an important part of the evaluation and comparison considerations, so as to gain operator acceptance.

The wastewater treatment plant is based on an anoxic/oxic model. This features an anoxic reactor, an aerobic reactor, and a clarifier aligned in series. The output concentrations of interest are effluent concentrations of carbonaceous waste (BOD), ammonia, and nitrate. The treatment objectives are to reduce the carbonaceous wastes and nitrogen compounds to water, carbon dioxide, and nitrogen gas. The input variables being considered to control the output variables are the recirculation ratio and the sludge age.

Two-by-two decentralized feedback controllers were designed for the plant. Two approaches are considered: a Proportional-Integral-Derivative (PID) based and a model based predictive control (MPC). Simulation results show that with the first level feedback control strategy of decoupled PID, long term stability issue is a big problem for those loops that exhibit delay times and inverse response characteristics. Additional compensatory control elements, such as lag filters, Smith predictors, and inverse response compensators would need to be added to improve the stability. However, controller complexity is also increased greatly, leading to increasingly complex tuning. Although, instability with pure PID controllers is manifested only very gradually (>100 days), its implications in application will be discussed in this report. On the other hand, decoupled model based control, the next level up in controller design hierarchy, such as the internal model control (IMC), can be easily tuned to yield long term stability of the closed-loop system, even when plant/model mismatch is present. However, the large and rapid control actions called for at times may not be feasible for implementation. In addition, the long time constant required in reaching a steady state may also not be feasible. In this case an advanced multivariable controller that can handle either input or output constraints explicitly would need to be considered. But this would be at a major technical and financial investment on the part of the industrial practitioner. If only a first level feedback controller is to be considered, then model based control appears to be more preferable to that of PID based because of its stability yielding characteristics, ease of tuning, and the ability to incorporate plant/model mismatch into the controller design. If a pure PID based controller is employed, very close monitoring and frequent returning of the controller is required in order to insure closed-loop stability. This need for constant vigilance may obviate any advantage of implementing an automated feedback controller over manual open-loop control by an experienced operator. Details of simulation results of the two approaches will be presented in this thesis.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS