Big data is a cornerstone component of the fourth industrial revolution, which calls onengineers and researchers to fully utilize data in order to make smart decisions and enhance the efficiency of industrial processes as well as control systems. In practice, industrial process control systems typically rely on a data-driven model (often linear) with parameters that are determined by industrial/simulation data. However, in some scenarios, such as in profit-critical or quality-critical control loops, first-principles concepts that are based on the underlying physico-chemical phenomena may also need to be employed in the modeling phase to improve data-based process models. Hence, process systems engineers still face significant challenges when it comes to modeling large-scale, complicated nonlinear processes. Modeling will continue to be crucial since process models are essential components of cutting-edge model-based control systems, such as model predictive control (MPC). Machine learning models have a lot of potential based on their success in numerousapplications. Specifically, recurrent neural network (RNN) models, designed to account for every input-output interconnection, have gained popularity in providing approximation of various highly nonlinear chemical processes to a desired accuracy. Although the training error of neural networks that are dense and fully-connected may often be made sufficiently small, their accuracy can be further improved by incorporating prior knowledge in the structure development of such machine learning models. Physics-based recurrent neural networks modeling has yielded more reliable machine learning models than traditional, fully black-box, machine learning modeling methods. Furthermore, the development of systematic and rigorous approaches to integrate such machine learning techniques into nonlinear model-based process control systems is only getting started. In particular, physics-based machine learning modeling techniques can be employed to derive more accurate and well-conditioned dynamic process models to be utilized in advanced control systems such as model predictive control. Along with Lyapunov-based stability constraints, this scheme has the potential to significantly improve process operational performance and dynamics. Hence, investigating the effectiveness of this control scheme under the various long-standing challenges in the field of process systems engineering such as incomplete state measurements, and noise and uncertainty is essential. Also, a theoretical framework for constructing and assessing the generalizability of this type of machine learning models to be utilized in model predictive control systems is lacking. In light of the aforementioned considerations, this dissertation addresses the incorporation ofprior process knowledge into machine learning models for model predictive control of nonlinear chemical processes. The motivation, background and outline of this dissertation are first presented. Then, the use of machine learning modeling techniques to construct two different data-driven state observers to compensate for incomplete process measurements is presented. The closed-loop stability under Lyapunov-based model predictive controllers is then addressed. Next, the development of process-structure-based machine learning models to approximate large, nonlinear chemical processes is presented, with the improvements yielded by this approach demonstrated via open-loop and closed-loop simulations. Subsequently, the reliability of process-structure-based machine learning models is investigated in the presence of different types of industrial noise. Two novel approaches are proposed to enhance the accuracy of machine learning models in the presence of noise. Lastly, a theoretical framework that connects the accuracy of an RNN model to its structure is presented, where an upper bound on a physics-based RNN model's generalization error is established. Nonlinear chemical process examples are numerically simulated or modeled in Aspen Plus Dynamics to illustrate the effectiveness and performance of the proposed control methods throughout the dissertation.

This book presents theoretical and practical findings on the state estimation, diagnosis and control of complex systems, especially in the mathematical form of descriptor systems. The research is fully motivated by real-world applications (i.e., Barcelona’s water distribution network), which require control systems capable of taking into account their specific features and the limits of operations in the presence of uncertainties stemming from modeling errors and component malfunctions. Accordingly, the book first introduces a complete set-based framework for explicitly describing the effects of uncertainties in the descriptor systems discussed. In turn, this set-based framework is used for state estimation and diagnosis. The book also presents a number of application results on economic model predictive control from actual water distribution networks and smart grids. Moreover, the book introduces a fault-tolerant control strategy based on virtual actuators and sensors for such systems in the descriptor form.

In this dissertation, we study the interaction between the control performance and the quality of the state estimation in a constrained Model Predictive Control (MPC) framework for systems with stochastic disturbances. This consists of three parts: (i) the development of a constrained MPC formulation that adapts to the quality of the state estimation via constraints; (ii) the application of such a control law in a multi-vehicle formation coordinated control problem in which each vehicle operates subject to a no-collision constraint posed by others' imperfect prediction computed from finite bit-rate, communicated data; (iii) the design of the predictors and the communication resource assignment problem that satisfy the performance requirement from Part (ii). Model Predictive Control (MPC) is of interest because it is one of the few control design methods which preserves standard design variables and yet handles constraints. MPC is normally posed as a full-state feedback control and is implemented in a certainty-equivalence fashion with best estimates of the states being used in place of the exact state. However, if the state constraints were handled in the same certainty-equivalence fashion, the resulting control law could drive the real state to violate the constraints frequently. Part (i) focuses on exploring the inclusion of state estimates into the constraints. It does this by applying constrained MPC to a system with stochastic disturbances. The stochastic nature of the problem requires re-posing the constraints in a probabilistic form. Using a gaussian assumption, the original problem is approximated by a standard deterministic constrained MPC problem or the conditional mean process of the state (the prediction). The state estimates' conditional covariances appear in tightening the constraints as measuring the necessary standoff from the bound on the real state. `Closed-loop covariance' is introduced to reduce the infeasibility and the conservativeness caused by using long-horizon, open-loop prediction covariances. The resulting control law is applied to a telecommunications network traffic control problem as an example. The idea of posing and transforming a probabilistic MPC problem works well, but not limited to, linear systems. In Part (ii), we consider applying constrained MPC as a local control law in a coordinated control problem of a group of distributed autonomous systems. Interactions between the systems are captured via constraints. First, we inspect the application of constrained MPC to a completely deterministic case. Formation stability theorems are derived for the subsystems and conditions on the local constraint set are derived in order to guarantee local stability or convergence to a target state. If these conditions are met for all subsystems, then this stability is inherited by the overall system. For the case when each subsystem suffers from disturbances in the dynamics, own self-measurement noises, and quantization errors on neighbors' information due to the finite-bit-rate channels, the constrained MPC strategy developed in Part (i) is appropriate to apply. Disturbance attenuation, or ``string stability", is studied in this framework and it is shown that inactivity of the MPC constraints implies stability. This then provides a connection between control objective, communications resource assignment and performance. A one-dimensional vehicle example is computed to crystallize ideas. The application of this part is not restricted to linear systems. In Part (iii), we discuss the local predictor design and bandwidth assignment problem in a coordinated vehicle formation context. The MPC controller used in Part (ii) relates the formation control performance and the information quality in the way that large standoff implies conservative performance. If the communication channels used to exchange local information are noiseless, but have only finite bit-rate, the bits assigned to each variable in the information package will change the prediction error covariance, and hence the control performance, via the quantization errors which can be regarded as measurement noises. In this part, we aim at deriving the minimal communication resource and the corresponding bit-rate assignment strategy the corresponding stable state predictors that is used to formulate the MPC constraints. We first develop an LMI (Linear Matrix Inequality) formulation for cross-estimator design in a simple two-vehicle scenario with non-standard information: one vehicle does not have access to the other's exact control value applied at each sampling time, but to its known, pre-computed, coupling linear feedback control law. Then a similar LMI problem is formulated for the bandwidth assignment problem that minimizes the total number of bits by adjusting the prediction gain matrices and the number of bits assigned to each variable. This LMI formulation takes care of the constraint on steady state prediction error covariance imposed by the formation performance requirement, the constraint on the limited total bandwidth, and the constraint on the predictors being stable. Some linear approximation is used to include the bandwidth assignment variables in the LMI formulation. The solution of the resulting LMIs guarantees the feasibility of the bandwidth assignment scheme and stable predictors, but not optimality. An example of a three-vehicle formation is also provided. The LMI formulation here is restricted to linear systems.

The high temperature solid oxide fuel cell (SOFC) is identified as one of the leading fuel cell technology contenders to capture the energy market in years to come. However, in order to operate as an efficient energy generating system, the SOFC requires an appropriate control system which in turn requires a detailed modelling of process dynamics. Introducting state-of-the-art dynamic modelling, estimation, and control of SOFC systems, this book presents original modelling methods and brand new results as developed by the authors. With comprehensive coverage and bringing together many aspects of SOFC technology, it considers dynamic modelling through first-principles and data-based approaches, and considers all aspects of control, including modelling, system identification, state estimation, conventional and advanced control. Key features: Discusses both planar and tubular SOFC, and detailed and simplified dynamic modelling for SOFC Systematically describes single model and distributed models from cell level to system level Provides parameters for all models developed for easy reference and reproducing of the results All theories are illustrated through vivid fuel cell application examples, such as state-of-the-art unscented Kalman filter, model predictive control, and system identification techniques to SOFC systems The tutorial approach makes it perfect for learning the fundamentals of chemical engineering, system identification, state estimation and process control. It is suitable for graduate students in chemical, mechanical, power, and electrical engineering, especially those in process control, process systems engineering, control systems, or fuel cells. It will also aid researchers who need a reminder of the basics as well as an overview of current techniques in the dynamic modelling and control of SOFC.