Full text loading...
-
Modular Neural Networks Reservoir Properties Prediction
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 4th EAGE St.Petersburg International Conference and Exhibition on Geosciences - New Discoveries through Integration of Geosciences, May 2010, cp-156-00048
- ISBN: 978-90-73781-79-5
Abstract
In this article we represent new technology which can improve performance of the reservoir properties prediction based on the seismic and well data, while using neural networks as prediction mechanism. This technology contains three main steps . 1-st step - Statistical or neural model based attributes selection and transformation. 2-nd step - Formal (algorithmic) or informal (geological) clustering of the source area. 3-rd step - Special (modular) learning of the set of neural networks and construction of the cooperative network to predict reservoir properties. Especially this work represent the use of modular, cooperative neural networks to improve generalization accuracy of quantitative interwell reservoir properties approximation using seismic attributes, well data and formal and informal data clustering techniques. There is also the real sample to show modular neural networks improved performance.