The research direction in the Multi-parametric Optimization & Control Group aims at gaining fundamental understanding, and developing theoretical and practical advances & tools in the areas of multi-parametric optimization and control, first-principles based and data-driven modeling, energy & process systems engineering, biological & biomedical systems engineering, sustainable smart manufacturing and process intensification. Current areas of interest include the energy-water-food nexus, circular economy issues, the interactions of energy - policy & law, big data analytics for environmental health decisions in emergency response, and resiliency enhancement in energy and manufacturing supply chains.
While the developments of multi-parametric programming theory and its application to MPC have allowed for the development of new and fundamentally different control paradigms, most real-world systems are too complex to be used in model-predictive control. Additionally, the different model-based tasks in process systems engineering such as parameter estimation, dynamic optimization, design, scheduling and control require different types of complexity and accuracy of the underlying model of the physical system.
In order to bridge this gap, we have developed PAROC, an integrated framework and software platform for the design and model-based control of process systems (see here for details). The basic idea is to decompose the problem of designing a model-based controller for a complex system into a series of steps: (i) 'high-fidelity' modelling and model analysis, (ii) system identification and model reduction techniques', (iii) application of multi-parametric programming to the approximate model, (iv) development of explicit state estimation techniques and (v) closed-loop validation of the developed controller 'in-silico' against the high-fidelity model.
One of the most important applications of multi-parametric programming is its use in model-predictive control (MPC), which enables the explicit solution of the resulting optimization problem. This allows the computational burden to be transferred offline, thus reducing the online computational effort to a point location and a function evaluation. Hence, it is possible to implement these optimal control actions to a standard micro-chip, hence coining the idea of 'MPC-on-a-chip'.
As the pioneers of this field and inventors of this technology (see patents EP1399784 and US7433743), we have been continuously developing connections between advances multi-parametric programming theory and explicit model predictive control in several areas such as dynamic programming, hybrid systems, continuous-time control, estimation and robust control. All of these contributions signify important developments and open possibilities in control that were previously unaccessible.
Developed in parallel to sensitivity analysis, the idea of solving optimization problems for a range and as a function of certain bounded parameters has gained considerable interest. Within our group, we have tackled this challenge by developing POP, the Parametric OPtimization toolbox. To date, it represents the most advanced software for multi-parametric programming, as it is the only software which includes reliable solvers for mp-MILP and mp-MIQP problems, as well as a comprehensive problem library of 400 test problems as well as a powerful problem generator.
The availability and use of energy is one of today's key challenges. With energy systems engineering, we provide a methodological, generic framework to arrive at realistic integrated solutions to complex energy and environmental problems. These problems existing in design, control and operation of energy systems require a holistic, system-based approach producing optimal design and operational plans for systems ranging from nanoscale, mirco-scale, mesoscale to mega-scale levels over horizons that range from milliseconds to months or years.
Within our group, our focus lies on superstructure-based modelling, mixed-integer linear and nonlinear programming, optimization and control under uncertainty and life-cycle assessment. These concepts are applied to systems such as the cogeneration of heat and power, and are integrated into the recently introduced PAROC framework and software platform.
Process intensification (PI) offers the potential to substantially improve chemical and manufacturing processes by realizing step changes in energy efficiency, cost reduction, and environment impact minimization through the development of novel process schemes and equipment. While many PI alternative technologies and their conventional counterparts exist, systematic approaches and tools to decide on the most promising intensified process solutions, simultaneously with operability and safety considerations, are currently rather lacking.
To address these challenges, we focus on the development of a holistic framework for the synthesis of operable process intensification systems leveraging process synthesis, optimization, and operability assessment methods. Key constituents of the proposed approach include: (i) a novel phenomena-based synthesis representation approach, i.e. the Generalized Modular Representation Framework, to discover intensified unit/flowsheet candidates, (ii) advanced control/operability/safety metrics to address the unique operational characteristics in intensified processes, and (iii) integration with the PAROC framework and software platform to ensure the delivery of verifiable and operable PI systems with explicit model predictive control strategies.
The advancement in technology and computational power has enabled huge amount of data collection in real time, which has initiated the big data era. Today big data analytics is becoming an essential tool for businesses as well as society, providing assistance in decision-making processes in numerous industries including energy, environment, and health.
Within our group, we focus on data-driven modeling, dimensionality reduction, and global optimization of constrained grey-box computational systems. Our interest lies in the development of data-driven nonlinear nonconvex optimization, multi-objective optimization, dynamic optimization, and bi-level optimization techniques, as well as machine learning-based dimensionality reduction methodologies. These theoretical advancements are applied to various fields of process systems engineering. Specifically, our group has established data-driven frameworks for integrated data-driven monitoring and explicit fault-aware control of chemical processes, food-energy-water nexus (FEW-N) decision making, predictive maintenance planning for smart manufacturing, oil-field operations, and chemicals production. Furthermore, we serve as the Data Science Core of the Texas A&M Superfund Research Center, where we build comprehensive models and tools for addressing exposure to unknown chemical mixtures during environmental emergency-related contamination events, such as hurricanes.
Data-Driven Tools for the Superfund Project
Due to population growth, economic development, urbanization, climate change, and natural resource degradation, the global demand for food, energy and clean/freshwater is rapidly increasing. As demand grows the stresses and interdependences between these three resource systems, which are commonly referred to as Food-Energy-Water Nexus (FEW-N), are becoming more apparent. Decision making can be very challenging due to multiple dimensions of the biophysical water, energy and food systems, and the players and stakeholders connected with them. A holistic systems engineering approach is thus clearly needed to navigate the multi-faceted FEW-N space, identify opportunities for synergistic benefits and systematically explore interactions and trade-offs.
To address this problem, we utilize a systems engineering framework that can systematically explore interactions and tradeoffs within FEW-N networks. The framework combines data analytics and mixed-integer modeling and optimization methods establishing the interdependencies and potentially competing interests amongst the food, energy and water elements in the system, along with policy, sustainability and feedback from the various stakeholders. A multi-objective optimization strategy is followed for the analysis of the trade-offs empowered by the introduction of composite FEW-N metrics as means to facilitate decision making and compare alternative process and technological options.
We are also part of the WEF Nexus Research Group and WEF Nexus Initiative (SACS-WEFNI), focusing on the science for planning the Water-Energy-Food Resources Nexus in San Antonio and surrounding regions as climate and urban growth alter water supplies.
Smart manufacturing (SM) encapsulates a new revolution in the manufacturing industry. It is envisioned to leverage network technologies such as the industrial Internet of things, big data analytics, smart assets (machinery or equipment) and highly skilled workforce to create a new manufacturing ecosystem that will make available the right data in the right form, the right people with the right knowledge, the right technology and the right operations, whenever and wherever needed. The smart manufacturing ecosystem seeks to aggregate various technology components and business models into a unified platform that will enable technology innovation, economic health and resources for an agile, safe and sustainable manufacturing industry. Towards the rapid development and deployment of smart manufacturing, a number of technologies are prioritized for accelerated research and development. The technology areas include; advances in sensing, control, platform, and modeling (ASCPM).
Our research group has been engaged in developing various elements of smart manufacturing in the last three decades. Within the PAROC framework and the MPC-on-a-Chip technology are; the development of various forms of models, model validation using advanced sensor technology and big data analytics, advanced control design and the development of a hardware interface for deploying the control algorithm.
Strategic decision making in the process industry is traditionally structured hierarchically, with an information flow dominantly from larger to smaller time scales. Independent and sequential assessment of these decision layers may lead to suboptimal, and even infeasible operations. Integration of these layers across an enterprise is expected to deliver more profitable and reliable operations by benefiting from the synergistic interactions between different decisions.
Within our group, we develop a unified theory and framework to integrate strategic decision-making based on a single high fidelity model. The framework features (i) a mixed-integer dynamic optimization (MIDO) formulation that accounts for the design, scheduling, and control decisions with safety and operability considerations, and (ii) a multiparametric optimization strategy for the derivation of offline/explicit maps of optimal receding horizon policies. The explicit expressions for the smaller time scale decisions allow for seamless and exact representation in the longer strategic decisions.
Energy affects every single individual and entity in the world. Therefore, it is crucial to precisely quantify the "price of energy" and study how it evolves through time, through major political, social and technological events, and through changes in energy and monetary policies. Hence, we have developed a predictive framework, the Energy Price Index (EPIC), to calculate the up-to-date average price of energy in the United States. The framework can be easily extended to calculate the average price of energy in other countries, states etc. The two key factors comprising EPIC are the total demand of the energy products that are directed to the end-use sectors along with their respective prices. The proposed framework enables the prediction of future energy demands and prices with high accuracy. Therefore, EPIC is the state-of-the-art for quantitatively evaluating, designing, testing, and optimizing various governmental policies, proposals, and questions related to energy.
Combined heat and power (CHP) systems constitute an upcoming technology that has the potential to replace the conventional processes used so far for the production of usable heat and electricity. The cogeneration of heat and power in a single process increases the system efficiency, while it decreases the operational cost. The use of CHP systems in the domestic/residential sector becomes appealing from (i) an environmental, (ii) economical and (iii) efficiency oriented point of view.
Aspects of high-fidelity modeling of such systems as well as their optimal design, operation and control are investigated via the application of the PAROC framework and the use or multi-parametric receding horizon techniques.
Fuel cells are a promising technology for electrical power generation, consisting of electrochemical devices that convert the chemical energy of a fuel to electrical energy. Its key characteristic is thereby its increased electrical efficiency as it avoids the intermediate generation of mechanical energy.
Within our group, we focus on the modelling and explicit model predictive control of a PEM fuel cell system. In particular, our group is equipped with a fully automated PEM fuel cell pilot plant, which allows us to validate our approaches on real life systems.
Fuel cells are a versatile source of power generation due to its efficient conversion of chemical to electrical energy of fuel. One natural application for such systems are cars, especially as fuel cells can be stacked into arrays, resulting in high energy-density systems.
Within our group, we have constructed a mini vehicle, which is powered by a fuel cell used to charge the battery. This energy is then used within the vehicle, which is stirred via an explicit model-predictive controller developed using the PAROC framework.