Adaptive water quality control in drinking water distribution networks

Regulating the spatio-temporal distribution of chlorine concentration is a crucial component of providing clean drinking water to consumers. Increasingly stringent requirements on the levels of chlorine residual and disinfectant by-products in drinking water distribution networks have created a need for feedback water quality control approaches. Based on investigating the spatially distributed input-output relationship of disinfectant residual in water distribution networks, this dissertation presents an input/output water quality modeling approach that is suitable for designing feedback control algorithms. The water quality control problem is formulated in an adaptive control framework, and a design approach based on parameter estimation and adaptive control scheme is proposed with special consideration on the periodic variation of parameter uncertainty due to varying consumer demands. The periodic parametric uncertainty is represented by a Fourier series with on-line parameter estimation of the unknown coefficients. The indirect adaptive control scheme for single-input single-output (SISO) distribution network is studied, and then is extended to the quality control with multi-booster stations by decomposing the network into sub-systems. A design and analysis procedure is developed for optimal control of chlorine maintenance for the case when the number of chlorine boosters (controllers) is less than the sensors. The general adaptive decentralized control scheme is investigated for control with multiple inputs based on the water quality modeling. The effectiveness of the proposed control schemes are illustrated via simulations based on a real water distribution network and an EPANET example network.