Neural Network Control of the Integrated Power System
Abstract
Neural networks are investigated for fault tolerant stabilization and control of an Integrated Power System (IPS). Neural networks can be robust in the sense that they are not disabled by incomplete or inconsistent information. As non-model based observers, neural networks are ideally suited to estimation of complex, interactive power systems. Specifically, the ability of neural networks to adapt to uncertain eventualities such as flooding, fire, and combat casualties is investigated. The IPS under consideration will provide integrated propulsion and ship's service power generation and distribution for the next generation of U.S. Navy surface ships also known as the DD-21. These solid state power systems involve nonlinear dynamics which can lead to negative impedance instability and voltage collapse. Feedforward back-propagating neural networks were evaluated with respect to variable structure and data degradation. This research represents an initial step toward unifying nonlinear, negative impedance stabilization with robust neural network fault detection and isolation. The Naval Sea Systems, Integrated Power System and the Office of Naval Research, Electrically Reconfigurable Ship programs motivated this research.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 07, 2000
- Accession Number
- ADA387932
Entities
People
- Johnathan J. Cerrito
Organizations
- United States Naval Academy