(544g) Deployment of a Search Space Reduction Algorithm to an NMPC Problem - Comparative Experimental Analysis to a Fuel Cell System | AIChE

(544g) Deployment of a Search Space Reduction Algorithm to an NMPC Problem - Comparative Experimental Analysis to a Fuel Cell System


Ziogou, C. - Presenter, Centre for Research and Technology-Hellas
Pistikopoulos, E. - Presenter, Texas A&M Energy Institute, Texas A&M University
Georgiadis, M. C. - Presenter, Imperial College
Voutetakis, S. - Presenter, C.P.E.R.I. / C.E.R.T.H.
Papadopoulou, S. - Presenter, Alexander Technological Educational Institute of Thessaloniki

Nowadays, the increase in energy demands, the tighter environmental regulations and various economic considerations require for the systems and processes to operate over a wider range of conditions and often near the boundaries of their operating region. Moreover, the interconnection and cooperation of various systems into integrated units is a prerequisite for efficient resource handling and allocation. In the sight of these issues, advance process control, such as Model Predictive Control (MPC) aims to facilitate the improvement of the response and overall behavior of systems and processes. In general the impact of control technology is evident in a wide range of application areas, including fuel cells, as it is the necessary facilitator for achieving desired objectives and fulfilling application-specific goals. Fuel cells represent a versatile and efficient electricity generation source that can be applied in a wide range of industries - from vehicles and primary energy systems to autonomous back-up power stations and portable consumer electronics devices. Fuel cell systems exhibit fast dynamics, nonlinearities and uncertainties that constitute challenges requiring appropriate control in order to be confronted effectively.

This work presents an alternative way of combining components from two advanced MPC methodologies for a multi-variable nonlinear control problem in order to systematically exploit their synergistic benefits. More specifically at the core of a nonlinear model predictive control (NMPC) formulation a nonlinear programming (NLP) problem is solved utilising a discretized nonlinear dynamic model. Prior to the online solution of the NLP problem a pre-processing search space reduction (SSR) algorithm is applied. The SSR algorithm adjusts the search space of selected variables, based on a pre-computed augmented low-complexity piecewise affine (PWA) approximation of the feasible space.

The behavior of the proposed synergetic formulation is illustrated through a multivariable nonlinear control problem involving the operation of a small scale fully automated Polymer Electrolyte Membrane (PEM) fuel cell unit. Overall, four different MPC-based approaches are formulated and subsequently evaluated by addressing the control issues that arise during the operation of the fuel cell system. The aim of comparative experimental study is to explore the behavior of MPC-based configurations and summarize the characteristics of each approach with respect to the specific control problem. To achieve this objective an experimental scenario is formed which is applied to the PEM fuel cell unit using each time a different control configuration. The proposed framework is developed and deployed online to an industrial-grade Supervisory Control and Data Acquisition (SCADA) system. The response of the multivariable nonlinear controllers are assessed through a set of experimental studies, illustrating that the control objectives are achieved and the fuel cell system operates at a stable environment regardless of the varying operating conditions. The performance of the deployed controllers is assessed based on their response, their qualitative response characteristics and the required energy for the heat-up and the cooling during each experiment. The comparative experimental study presents the benefits and challenges of each configuration.