Advances in predictive control and feedback equilibrium-seeking with applications to autonomous water resource recovery

Abstract

A paradigm shift is underway: Wastewater, historically viewed as an environmental hazard, is now recognized as a sustainable source of clean water, energy, and nutrients. Under this context, wastewater treatment plants (WWTPs) are being rethought as water resource recovery facilities (WRRFs), plants that produce goods and energy using wastewater as a raw material. While current efforts focus on designing novel processes, this dissertation considers the role of automatic decisionmaking (that is, feedback control) in this transition instead. We propose two research directions: (i) The repurposing of already existing infrastructure to these emerging objectives, and (ii) The promotion of a wastewater-centered market through an interconnected network of WRRFs.

In the first direction, the thesis studies the predictive control of conventional biological WWTPs for water resource recovery tasks. The proposal consists of an output-feedback control framework comprised of an operating point optimizer, a model predictive controller, and a moving horizon estimator. The framework is designed to control a WWTP to recover nutrients directly into effluent streams, thus producing reused water of tailored quality, while ensuring that such an operation regime can be fully sustained by the energy recovered from processing sludge into biogas. Using well-established benchmark models, experimental results demonstrate the efficacy of this control strategy in operating a conventional WWTP as an energy-autonomous WRRF.

In the second direction, the thesis studies the control of (noncooperative) multi-agent systems. The underlying idea is to connect WRRFs into a supply chain exchanging wastewater and recovered resources. Here, the contribution is theoretical: it addresses the open problem of seeking gametheoretical equilibria (e.g., the Nash equilibrium) of output-feedback policies that incorporate both operational and informational constraints. The thesis proposes novel equilibrium-seeking algorithms in which noncooperative agents are able to converge to an equilibrium of feedback policies, while simultaneously operating their subsystem under the aforementioned constraints. We provide formal convergence certificates and demonstrate the algorithms in exemplary problems.

Our theoretical and experimental findings contribute a step towards the transition into zerowaste water resource recovery infrastructures. In turn, the contributions to multi-agent control should benefit applications in different domains, such as smart grids and supply chain management.