Operational optimization and realtime control of fuelcell systems
Abstract
Fuel cells is a rapidly evolving technology with applications in many industries including transportation, and both portable and stationary power generation. The viability, efficiency and robustness of fuelcell systems depend strongly on optimization and control of their operation. This paper presents the development of an integrated optimization and control tool for Proton Exchange Membrane FuelCell (PEMFC) systems. Using a detailed simulation model, a database is generated first, which contains steadystate values of the manipulated and controlled variables over the full operational range of the fuelcell system. In a second step, the database is utilized for producing Radial Basis Function (RBF) neural network "metamodels". In the third step, a NonLinear Programming Problem (NLP) is formulated, that takes into account the constraints and limitations of the system and minimizes the consumption of hydrogen, for a given value of power demand. Based on the formulation and solution of the NLP problem, a lookup table is developed, containing the optimal values of the system variables for any possible value of power demand. In the last step, a Model Predictive Control (MPC) methodology is designed, for the optimal control of the system response to successive seppoint changes of power demand. The efficiency of the produced MPC system is illustrated through a number of simulations, which show that a successful dynamic closedloop behaviour can be achieved, while at the same time the consumption of hydrogen is minimized.
 Publication:

Journal of Power Sources
 Pub Date:
 2009
 DOI:
 10.1016/j.jpowsour.2009.01.048
 Bibcode:
 2009JPS...193..258H
 Keywords:

 3D;
 ThreeDimensional;
 CPU;
 Central Processor Unit;
 DC;
 Direct Current;
 FIR;
 finite impulse response;
 FC;
 fuel cell;
 LOO;
 Leave OneOut;
 LQG;
 Linear Quadratic Gaussian;
 LQR;
 Linear Quadratic Regulator;
 MIMO;
 multiinputmultioutput;
 MPC;
 Model Predictive Control;
 NLP;
 NonLinear Programming;
 NNM;
 neural network model;
 PEMFC;
 Proton Exchange Membrane FuelCell;
 PID;
 proportionalintegralderivative;
 PRESS;
 Prediction Error Sum of Squares;
 RBF;
 Radial Basis Function;
 RMSE;
 Root Mean Squared Error;
 SOFC;
 Solid Oxide Fuel Cell;
 SSE;
 Sum of Squared Errors between the observations and the predicted values;
 SSY;
 Sum of Squared Deviations between the observations and their mean