Please use this identifier to cite or link to this item: http://hdl.handle.net/11023/2085
Title: ONLINE IDENTIFICATION AND ADAPTIVE CONTROL OF BIOLOGICAL WASTEWATER TREATMENT PROCESS
Author: SADEGHASSADI, MAHSA
Advisor: Macnab, Chris J.B
Westwick, David
Keywords: Engineering--Electronics and Electrical
Issue Date: 5-Feb-2015
Abstract: The activated sludge process is a nonlinear, time-varying, multi-input-multi-output system. It is important to design an adaptive control strategy, which can not only track the setpoint changes quickly, but also adapt to the system uncertainties and disturbance signals. The current work presents a single-input-single-output Generalized Predictive Controller to regulate the dissolved oxygen concentration and a multi-loop Generalized Predictive Control scheme to control the dissolved oxygen concentration and the substrate concentration. The control is tested with both a simplified four-state model and an activated sludge model no. 1 (ASM1) benchmark of an activated sludge process. In the univariate algorithm, the manipulated variable (controller output) is the air flow rate (or the oxygen transfer coefficient) and the controlled variable (measured variable) is the dissolved oxygen concentration. On the other hand, for the multivariable process, two separate control loops are taken into consideration such that in the first control loop, the air flow rate (or the oxygen transfer coefficient) is used to regulate the dissolved oxygen concentration, and in the second control loop, the recycled to influent rate (or the return sludge flow rate) is used to manipulate the substrate concentration. Initially, the GPC controller is implemented in the presence of white measurement noise. In this case, an ARX model is used as the prediction model in a GPC structure. The RLS identification method estimates the parameters of this ARX model. Then, the ability of GPC controller in the presence of random walk disturbance (coloured noise together with an integral factor) is tested. In this case, a CARIMA model is used as the prediction model and an ELS identification method is used to estimate the parameters of this model. A method which can compensate the effect of coloured noise is GPC with T filter. The results of a conventional GPC controller is compared with a GPC with T filter and a PI controller.
URI: http://hdl.handle.net/11023/2085
Appears in Collections:Electronic Theses

Files in This Item:
File Description SizeFormat 
ucalgary_2015_Sadeghassadi_Mahsa.pdf6.8 MBAdobe PDFView/Open


Items in The Vault are protected by copyright, with all rights reserved, unless otherwise indicated.