ECONOMIC BASED CONTROL SYSTEM DESIGN
Omell, Benjamin Peter
EMPC differs from traditional MPC by directly utilizing a profit based function as the objective as opposed to a quadratic function that minimizes the distance from a set point that is predetermined. However, implementation of EMPC can result in unexpected and at times pathological closed-loop behavior, including inventory creep, bang-bang actuation and instability. To address these issues, an infinite-horizon version of EMPC is developed and shown to avoid many of the performance issues observed in the finite-horizon version. First, modifications to the EMPC problem will be used for the conceptual development of the Economic Linear Optimal Controller (ELOC), which is a statistically constrained linear feedback controller. Then, pointwise- in-time constraints can be reintroduced using one of two methods; Constrained ELOC or Infinite-Horizon EMPC (IH-EMPC). We also investigate the impact of problem formulation modifications on the ELOC. The first issue is that of disturbance modeling and the second is the impact of controller sample-time. The third topic concerns incorporation of computational delay in the feedback-loop, using both full and partial state information structures. Finally an illustration of the impact of plant-model mismatch is presented. The Constrained ELOC formulation is further modified to allow for market responsive smart grid applications. In particular an Integrated Gasification Combined Cycle (IGCC) process with hydrogen storage will be used to demonstrate the Constrained ELOC for such applications. The ELOC will be used as a vehicle to exploit dispatch capabilities by pursuing directly the objective of maximizing revenue. The idea being that process modifications to enable dispatch capabilities will allow for a time-shift of power production away from periods of low energy value to periods of high value. An in depth discussion is provided on how energy value forecasts are incorporated into the design of the constrained ELOC. Finally, an extension of the ix ELOC to the controller embedded equipment design is provided. The work concludes with a discussion of the computational aspects of solving the ELOC problem. In particular, the impact of reverse-convex constraints inherent to the ELOC problem are discussed along with existing solution methods. The main contribution of this final chapter is a novel application of the Generalized Benderâ€™s Decomposition (GBD) algorithm to the ELOC problem. This new approach is shown to retain global optimality, reduce computational effort (by orders of magnitude) and expand the class of problems one can solve.
PH.D in Chemical Engineering, December 2013
Chmielewski, Donald
2013
2013-12
Dissertation
application/pdf
islandora:9006
http://hdl.handle.net/10560/3260
ChBE / Chemical and Biological Engineering
Illinois Institute of Technology
en
In Copyright
http://rightsstatements.org/page/InC/1.0/
Restricted Access