An optimal reference governor with a neural network combined model for hybrid Fuel-Cell/Gas-Turbine
ABSTRACT This paper introduces a concept of real-time optimization of hybrid fuel-cell power plants as an alternative distributed generation source that improves the power quality and reliability of the power grid. One of the most important issues of plant operation is the optimal control of the power plant, leading to significant economic and environmental benefits. As a commercialized fuel cell technology, Direct Fuel-Cell with Gas-Turbine (DFC/T) power plant is investigated in this paper. A framework of an optimal reference governor (ORG) is developed to generate optimal control strategies for the local controllers. For the purpose of on-line application, a neural network combined model is built as a state estimator that approximates the plant behaviors, which is compatible with population based real-time heuristic optimization algorithms. The simulation of the optimization result is presented and validated by a comparison with experimental data and simulation result of a mathematical plant model.
Conference Proceeding: Operation and control of direct reforming fuel cell power plant[show abstract] [hide abstract]
ABSTRACT: Computer simulation is used to analyze the operation and efficiency of a carbonate fuel cell power plant under load perturbations. The plant model is based on a 2 MW system design used in the Santa Clara Demonstration Project and includes: internal reforming carbonate fuel cell stack, cathode gas preparation system, heat recovery unit and fuel processing system. Model development for various processes is based on thermochemical principles and conservation of mass and energy. Overall plant efficiency is determined by net fuel consumption based on calculated gas compositions and auxiliary power consumption. During load maneuvering, several key operational constraints must be maintained. Among these are: allowable stack temperature deviation, baseline fuel utilization, steam/carbon ratio, and pressure difference between anode and cathode. Actual plant control schemes are used in the simulation and are evaluated for performance under load changes. The results of these simulations will be used as a benchmark and development tool for advanced intelligent controllers for autonomous and efficient operation of fuel cell systemsPower Engineering Society Winter Meeting, 2000. IEEE; 02/2000
- [show abstract] [hide abstract]
ABSTRACT: A nonlinear programming (NLP) framework is developed to determine optimal operating policies for hybrid fuel cell/gas turbine power systems. The approach consists of a dynamic model of the power plant, reformulated as an index one differential algebraic equation (DAE) system. A dynamic optimization framework is developed where the constraints include the dynamic model of the plant. The system model is then discretized using Radau collocation on finite elements and formulated in the AMPL modeling environment. This allows for the straightforward solution of dynamic optimization problems using large-scale NLP solvers. IPOPT is the NLP solver used in this study. Program links were provided to Matlab/Simulink to visualize and interpret the results. The formulation of a dynamic optimization problem was focused on determination of optimal operating trajectories for tracking power plant load variations. Efficiency measures were also included as a part of the dynamic optimization problem to maximize efficiency while tracking the desired load profile. Results from 18 case studies show that the dynamic optimization can be performed quickly with excellent results. The applicability of the dynamic optimization framework for the estimation of feed fuel concentrations is also demonstrated. © 2007 American Institute of Chemical Engineers AIChE J, 2007AIChE Journal 01/2007; 53(2):460 - 474. · 2.49 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: This paper constitutes a first study of the Particle Swarm Optimization (PSO) method in Multiobjective Optimization (MO) problems. The ability of PSO to detect Pareto Optimal points and capture the shape of the Pareto Front is studied through experiments on well--known non--trivial test functions. The Weighted Aggregation technique with fixed or adaptive weights is considered. Furthermore, critical aspects of the VEGA approach for Multiobjective Optimization using Genetic Algorithms are adapted to the PSO framework in order to develop a multi--swarm PSO that can cope effectively with MO problems. Conclusions are derived and ideas for further research are proposed.04/2002;