Robust model predictive control of water quality in drinking water distribution systems
This thesis develops online feedback control of chlorine residuals performing at the lower level
of a hierarchical structure of integrated quantity and quality control in drinking water distribution
systems (DWDS), which provides a practical solution for online water quality control in DWDS.
Input-output and state-space models of the chlorine residuals are developed from mathematical
models of chlorine residual dynamics. The existing path analysis algorithm is extended and
utilized to obtain the parameter structure. Joint parameter and model structure error estimation is
developed using bounding approach based on a point-parametric model. The uncertainty radius of
the system is outlined through robust output prediction, through which requirements for model
accuracy from robust model predictive control (MFC) are explicitly imposed on model
estimation. Hence, an integrated design of controller and model estimation is achieved.
MFC is applied for chlorine residual control based on the set-bounded model. To fulfil output
constraints under system uncertainties, safety zones are employed, which are designed from an
online evaluation of the uncertainty scenarios of the system, to restrict the output constraints. The
safety zones can be obtained by solving a nonlinear constrained optimization problem using a
significantly simplified relaxation-gain algorithm. The resulting robust MFC (RMPC) is
decentralized assuming communication among the decentralized RMPCs is available.
The proposed methodology is verified by applying it to a simulated benchmark DWDS.
Simulation study of model estimation and RMPC operation is presented and discussed.