The increasing population and adverse climate conditions are escalating fresh water scarcity globally. Since irrigation consumes a large portion of fresh water, it is very important to improve irrigation efficiency. One such method on improving irrigation efficiency is to use a closed-loop scheme instead of the traditional open-loop irrigation schemes. In this thesis, a systems engineering approach is used to optimally place sensors for agro-hydrological systems, and to study the impacts of closed-loop scheduling and control in irrigation water management.

The contributions of this thesis include the development of distributed state estimation methods for nonlinear process networks, the development of systematic approaches to properly decompose general nonlinear processes into subsystems for distributed estimation, and the application of the developed methods in different processes and output-feedback fault detection and isolation.

In this thesis, we propose a computationally efficient economic model predictive control (EMPC) design which is based on a well-known methodology - the separation of the control and prediction horizon. The extension of the prediction horizon of EMPC is realized by employing an auxiliary control law which asymptotically stabilizes the optimal steady state. The contributions of this thesis are to systematically analyze the stability and performance of the general EMPC scheme with extended horizon, and to explore its extensions and applications to several specific scenarios.

In this thesis, we propose a distributed adaptive high-gain extended Kalman filtering approach for nonlinear systems. Specifically, we consider a class of nonlinear systems that are composed of several subsystems interacting with each other via their states. In the proposed approach, an adaptive high-gain extended Kalman filter is designed for each subsystem. The distributed filters communicate with each other to exchange subsystems' estimates. First, assuming continuous communication among the distributed filters, an implementation strategy which specifies how the distributed filters should communicate is designed and the detailed design of the subsystem filter is described. Second, we consider the case where the subsystem filters communicate to exchange information at discrete-time instants. Following this, the problem of time-varying delays and data losses in communications between subsystems' estimators is considered. For these two latter cases, a state predictor is used in each subsystem filter to provide predictions of the states of other subsystems. Also, to reduce the number of information transmission among the filters and prevent data trafficking, a triggered communication strategy is developed. The stability properties of the proposed distributed estimation schemes with the described communication types are analyzed. Finally, the effectiveness and applicability of the proposed schemes are illustrated via the applications to simulated chemical processes and a Three-Tank experimental system.

Chemical and petrochemical plants typically integrate a number of geographically distributed operating units, which are physically linked through energy and material streams or inherently coupled via plant-wide constraints. The main drawback of the current decentralized control system is that it fails to consider the interrelations between subsystems, which could usually result in poor performance or even loss of closed-loop stability. Such concerns have motivated various control strategies to tackle these problems. One possibility is to replace the whole network with a centralized control structure. Despite the potential benfits, this renovation would require significant capital cost, increase maintenance costs, and reduce fault tolerance. Another practical approach is a distributed control that aims to improve the performance of a currently installed decentralized network. Distributed model predictive control (DMPC) methods are divide into two general categories: non-coordinated and coordinated schemes. Coordinated DMPC (CDMPC) networks, which consist of distributed controllers and a coordinator, are able to attain an overall optimal solution over a wide range of conditions. The focus of this thesis is to develop on-line strategies for CDMPC systems and overcome existing issues with global convergence and stability of closed-loop systems, under price-driven CDMPC concept. In particular, the main contributions are developing two novel information flow mechanisms for CDMPC of nonlinear systems and proposing a new solution method for CDMPC of linear systems, via a bi-level optimization framework.

This thesis focuses on improving the computational efficiency of EMPC for WWTPs. Two distributed EMPC designs are presented in this thesis. In the first design, the centralized model is used in each subsystem EMPC controller design; and in the second design, a subsystem model is used in each subsystem EMPC design. The performance of these two distributed EMPC designs are compared with a centralized model predictive control (MPC) scheme and a centralized EMPC scheme from different aspects including effluent quality, operating cost, and computational efficiency. Model reduction methods are also applied to the WWTP process in this thesis. In particular, the trajectory piecewise linear model and the order reduced trajectory piecewise linear model are used to approximate the original nonlinear system model. Two EMPC designs are proposed based on the two models. The approximated model accuracy are compared with the original nonlinear model. The performance of these two EMPC designs are compared with the EMPC based on the nonlinear model from control performance and computational efficiency points of view. We also investigate how the number of linearization points affect the EMPC control performance and computational efficiency through these applications.

Chronic Kidney Disease (CKD) affects millions of people throughout the world today. One of the major side effects of this disease is the inability to regulate the body's red blood cell production, and subsequently the mass of the protein called hemoglobin within the body. The health of these patients deteriorates and they become anemic. In this thesis, the objectives are (1) to present an artificial patient simulator developed exclusively based on measurement noise and time-varying parameters in Pharmacokinetics and Pharmacodynamics (PKPD) model, (2) performance assessment of nonlinear constrained ARX model (C-ARX), and (3) hemoglobin modeling technique with modified constrained ARX modeling (C-ARX) method utilizing additional measurment such iron saturation and white blood cell (WBC) count. The hemoglobin response modeling methods are compared on a clinical dataset containing 167 patients. It demonstrates that the new modeling method offers better modeling results to the previously developed C-ARX model.

Chronic Kidney Disease (CKD) affects millions of people throughout the world today. One of the major side effects of this disease is the inability to regulate the body's red blood cell production, and subsequently the mass of the protein called hemoglobin within the body. The health of these patients deteriorates and they become anemic. This thesis presents several different hemoglobin response modeling techniques including classical ARX, pharmacokinetic and pharmacodynamic (PKPD) delayed differential equation modeling and a novel new nonlinear constrained ARX modeling (C-ARX) method. Different model predictive controllers were tested against the current anemia management protocol (AMP) from a participating hospital. The final controller recommended for use is a weighted recursive least squares zone model predictive controller that uses a funnel shaped control zone.

This thesis focuses on the development of coordinated distributed state estimation schemes. Specially, we propose coordination algorithms for distributed moving horizon state estimators (MHEs) for discrete-time linear systems. In particular, the class of linear system is composed of several subsystems that interact with each other via their states. Two coordination algorithms are studied: the price-driven coordination algorithm and the prediction-driven coordination algorithm. In the proposed coordinated distributed MHE (CDMHE) schemes, each subsystem is associated with a local MHE. In the design of a local MHE, a coordinating term is incorporated into its cost function which is determined by an upper-layer coordinator. It is shown that both CDMHE schemes are able to achieve the estimation performance of the corresponding centralized design if convergence at each sampling time is ensured.

The processing of coal in coal handling and preparation plants (CHPPs) produces a significant amount of valuable fine coal which is then recovered using froth flotation. As froth flotation units are regulated using simple, conventional control methods the application of modern control methods is one option to improve the efficiency and profitability of these units. There are various issues to consider before modern control methods can be applied, namely: state estimation, handling model plant mismatch (MPM) and handling measurement delays. This thesis compares the performance of various modern control methods in the nominal case (no MPM and no measurement delays) and in the case of MPM with no measurement delays with the aim of identifying the best control method for each scenario. State estimation methods are also tested in the nominal case, with MPM and with measurement delays to identify the best one for each scenario.

This thesis presents a robust distributed moving horizon state estimation (DMHE) scheme that is appropriate for output feedback distributed predictive control of nonlinear systems as well as approaches for reducing the communication demand of the proposed DMHE scheme and a strategy for handling delays in the communication between subsystem estimators. First, the proposed robust DMHE scheme is presented for a class of nonlinear systems that are composed of several subsystems. It is assumed that the subsystems interact with each other via their states only. Subsequently, two triggered communication algorithms are introduced for the proposed DMHE scheme to reduce the number of information transmissions between subsystems. Following this, an approach is proposed to handle the potential time-varying delays in the communication between the subsystem estimators. The applicability and effectiveness of the proposed approaches are illustrated via their applications to different chemical process examples.

The focus of this thesis is the computational properties, mainly convergence and computational complexity, of two linear coordinated distributed model predictive control (DMPC) schemes: prediction-driven coordinated DMPC and price-driven coordinated DMPC. First, by restricting the study to linear unconstrained systems, the DMPC algorithms are transformed into iterative forms. Subsequently, explicit expressions for their convergence accuracy, convergence rates and the computational complexities are derived. A series of numerical experiments were also conducted to study the two DMPC methods' empirical computational complexity. It was discovered that DMPC methods' computational load is closely related to the local MPC design as well as factors like size and number of subsystems.