Abstract :
[en] Humanity has experienced a drastic growth during the last decades, faster than what was experienced in previous centuries. Increment concerns not only the population number but the amount of developments in areas such as computation capacity, automation, robotics, precision agriculture, material science, among others. These innovative solutions bring together a higher energy demand, hence in order to diminish the green house effect, global warming and other impacts caused by industry and oil based transportation system to the natural resources, it is deemed necessary to achieve sustainable and more efficient manufacturing processes.
Regarding energy efficiency organic Rankine cycle (ORC) power systems appear as an interesting technology to recover waste heat available at low-grade temperature, thus increasing the overall system efficiency. The ORC unit operates under the same principle as classical Rankine cycle for electricity generation, the main difference being the use of refrigerants with low boiling point as working fluid instead of water, allowing to reach superheated state for low temperature heat source conditions, thus becoming ideal for waste heat recovery (WHR) applications in the range from 100 400 C. The optimal thermodynamic design of such machines represents a first challenge, since cycle architecture, components and refrigerant selection, are an important aspect to look at. However such decisions are often considered at steady-state design, while ORC operates outside the designed range due to varying waste heat profiles, thus making necessary to consider a suitable control strategy aiming to guarantee safety operation and optimal
performance during transient conditions. The main contributions of this PhD thesis include: experimental validation of the proposed strategies, open-loop tests to illustrate the dynamics that represent a real challenge for modeling and control, presenting the conditions to achieve optimal ORC operation by building an optimizer, defining gain-scheduling and adaptive strategies based on classical PID and more advanced Model Predictive Control (MPC) to deal with nonlinear time-varying ORC dynamics. In order to
provide additional robustness against modeling errors a multiple model predictive controller is designed, where a Bayesian weighting scheme is applied to obtain an average prediction trajectory from a model bank built with models identified at different operating points. In order to better explain the complex ORC dynamics an sparse identification algorithm is proposed aiming to built a global nonlinear description of the process.
This research work is thus an attempt to present a user-friendly methodology for waste heat recovery organic Rankine cycle (WHR-ORC) modeling and control trol, a guide for practitioners and researchers interested on understanding from a data-driven perspective why is this power unit a nonlinear time-varying system, how to define a suitable low-order model for control, and how to design an advanced control strategy to achieve optimal performance under drastic waste heat variations. As well as to provide some ideas and future perspectives to optimize the ORC performance.
