Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

U-tube steam generator empirical model development and validation using neural networks

Conference · · Transactions of the American Nuclear Society; (United States)
OSTI ID:7193459
;  [1];  [2]
  1. Texas A and M Univ., College Station (United States)
  2. Dynamica, Inc., Houston, TX (United States)

Empirical modeling techniques that use model structures motivated from neural networks research have proven effective in identifying complex process dynamics. A recurrent multilayer perception (RMLP) network was developed as a nonlinear state-space model structure along with a static learning algorithm for estimating the parameter associated with it. The methods developed were demonstrated by identifying two submodels of a U-tube steam generator (UTSG), each valid around an operating power level. A significant drawback of this approach is the long off-line training times required for the development of even a simplified model of a UTSG. Subsequently, a dynamic gradient descent-based learning algorithm was developed as an accelerated alternative to train an RMLP network for use in empirical modeling of power plants. The two main advantages of this learning algorithm are its ability to consider past error gradient information for future use and the two forward passes associated with its implementation. The enhanced learning capabilities provided by the dynamic gradient descent-based learning algorithm were demonstrated via the case study of a simple steam boiler power plant. In this paper, the dynamic gradient descent-based learning algorithm is used for the development and validation of a complete UTSG empirical model.

OSTI ID:
7193459
Report Number(s):
CONF-920606--
Journal Information:
Transactions of the American Nuclear Society; (United States), Journal Name: Transactions of the American Nuclear Society; (United States) Vol. 65; ISSN TANSA; ISSN 0003-018X
Country of Publication:
United States
Language:
English