Dam management with imperfect models: bayesian model averaging and neural network control

Darwen, Paul J. (2012) Dam management with imperfect models: bayesian model averaging and neural network control. In: Proceedings of 8th International Conference on Intelligent Computing (304) pp. 360-366. From: ICIC 2012, 25-29 July 2012, Huangshan, China.

Preview

PDF (Submitted Version) - Submitted Version
Download (2MB)

PDF (Published Version) - Published Version
Restricted to Repository staff only

DOI: 10.1007/978-3-642-31837-5_52

View at Publisher Website: http://dx.doi.org/10.1007/978-3-642-3183...

1213

Abstract

Dam management is a controversial control problem for two reasons. Firstly, models are (by definition) crudely simplified versions of reality. Secondly, historical rainfall data is limited and noisy. As a result, there is no agreement on the "best" control policy for running a dam. Bayesian model averaging is theoretically a good way to cope with these difficulties, but in practice it degrades under two approximations: discretizing the parameter space, and excluding models with a low probability of being correct. This paper explores the practical aspects of how Bayesian model averaging with a neural network controller can improve dam management and flood control.


Item ID:	22364
Item Type:	Conference Item (Research - E1)
ISBN:	978-3-642-31836-8
Related URLs:	Author http://www.ic-ic.org/2012/
Date Deposited:	21 Jan 2013 22:31
FoR Codes:	08 INFORMATION AND COMPUTING SCIENCES > 0801 Artificial Intelligence and Image Processing > 080108 Neural, Evolutionary and Fuzzy Computation @ 80% 08 INFORMATION AND COMPUTING SCIENCES > 0801 Artificial Intelligence and Image Processing > 080110 Simulation and Modelling @ 20%
SEO Codes:	89 INFORMATION AND COMMUNICATION SERVICES > 8902 Computer Software and Services > 890201 Application Software Packages (excl. Computer Games) @ 80% 96 ENVIRONMENT > 9609 Land and Water Management > 960913 Water Allocation and Quantification @ 20%
Downloads:	Total: 1213 Last 12 Months: 10
	More Statistics

Actions (Repository Staff Only)

Item Control Page