A Hybrid Intelligent System For Improved Petrophysical Predictions
Neural networks have shown high potential for solving highly non-linear problems. In many instances, the success of the application is highly dependent on the ability to quantify the prediction uncertainty and the availability of a good training set. Bayesian neural networks provide a promising tool in this area. In this paper, we propose to incorporate an improved data selection strategy for Bayesian networks. The improved strategy includes the use of decision tree for the removal of less relevant input variables and the use of Bayesian error bar for pattern selection. The new training set is used to train the Bayesian networks. The case study from an onshore oilfield data set from west China shows that the proposed system gives significant improvement in a blind test and produces more reliable and accurate predictions.
Bayesian neural networks decision tree error bar petroleum reservoir permeability.
D. Qu P.M. Wong S. Cho T.D. Gedeon
School of Petroleum Engineering, University of New South Wales, Sydney, Australia Department of Industrial Engineering, Seoul National University, Seoul, Korea School of Information Technology, Murdoch University, Perth, Australia
国际会议
8th International Conference on Neural Information Processing(ICONIP 2001)(第八届国际神经信息处理大会)
上海
英文
1610-1614
2001-11-14（万方平台首次上网日期，不代表论文的发表时间）