Intelligent machine learning analysis for phosphoric acid fuel cell operations
Several fuel cell types are available and are in various stages of technology development. The complex nature of the balance of plant and fuel cell interface poses many technical challenges to achieve proper system control under commercial operating conditions. Real-time predictive diagnostic computer systems based on advanced intelligent machine learning technologies offer a means to facilitate the detection, understanding, and control of fuel cell subsystems to avoid system instabilities and failures that can result in costly plant shutdowns. The objectives reported herein are the development of physical and empirical computer models for application and testing of predictive control strategies based on intelligent machine learning techniques for fuel cells. A physical/empirical model was built and validated using available operating data from commercial fuel cells. Neural networks were then used to build an empirical model from the original physical/empirical model. Using the neural network model, a predictive, feedforward strategy was developed to control the fuel flow for a phosphoric acid fuel cell physical/empirical model. The predictive control strategy was compared to traditional proportional integral derivative control schemes.
- Research Organization:
- ERC, Inc., Tullahoma, TN (US)
- OSTI ID:
- 20002752
- Resource Relation:
- Conference: 33rd Intersociety Energy Conversion Engineering Conference, Colorado Springs, CO (US), 08/02/1998--08/06/1998; Other Information: 1 CD-ROM. Operating system required: Windows 3.x; Windows 95/NT; Macintosh; UNIX. All systems need 2X CD-ROM drive.; PBD: 1998; Related Information: In: Proceedings of the 33. intersociety energy conversion engineering conference, by Anghaie, S. [ed.], [2800] pages.
- Country of Publication:
- United States
- Language:
- English
Similar Records
IEA/AIE-89; Proceedings of the Second International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems, University of Tennessee, Tullahoma, June 6-9, 1989. Volumes 1 2
Neural networks: A versatile tool from artificial intelligence