Confidence Estimation for Petrophysical Predictions: An Intelligent Approach

Obtaining a permeability profile from well logs in heterogeneous reservoirs is often a complex and laborious process. The use of intelligent computing for permeability prediction has removed many of these difficulties. Conventional neural network models provide only an average error for the entire training set. This paper presents a simple hybrid technique for estimating confidence bounds for each permeability prediction. The case study in a clastic reservoir, offshore China, clearly demonstrates the usefulness of the technique and it gives more realistic results than those obtained from a conventional neural network. It is also simple to implement in conventional spreadsheets.