Shabnam Shahbazi; Abdorrahim Javaherian; Mojtaba Mohammadoo Khorasani
Abstract
Geological facies interpretation is essential for reservoir studying. The method of classification and identification seismic traces is a powerful approach for geological facies classification and distinction. Use of neural networks as classifiers is increasing in different sciences like seismic. They ...
Read More
Geological facies interpretation is essential for reservoir studying. The method of classification and identification seismic traces is a powerful approach for geological facies classification and distinction. Use of neural networks as classifiers is increasing in different sciences like seismic. They are computer efficient and ideal for patterns identification. They can simply learn new algorithms and handle the nonlinearity of seismic data. They are often reliable with noisy data or atypical environments. In this paper, an approach is presented based on competitive neural network for classification and identification of the reservoir facies that uses seismic trace shape. The competitive networks can be applied on discrete facies. Its unsupervised methods are independent on the wells data and other auxiliary information. Its supervised methods are independent on the wells location. This approach can be performed in two ways. In first way, the seismic facies are classified based on entirely on the characteristics of the seismic responses, without requiring the use of any well information. It is implemented by a single layer competitive unsupervised neural network, called Kohonen self organized neural network. In the second way, automatic identification and labeling of the facies is performed by the use of seismic responses and wells data. It is implemented by a two layer competitive supervised neural network, called Learning Vector Quantizer (LVQ) neural network. The results of both analyses on artificial seismic section and actual seismic section of the sixth zone of Asmary formation in Shadegan oilfield showed reservoir facies distribution and predicted heterogeneity of their characteristics.