Data Mining Well Logs for Optimum Well Placement

Well placement with geosteering can get very complex in reservoirs with formation change not simply addressed by changes in the gamma ray log response. This paper uses data mining to characterize complex reservoirs for optimum well placement. The objective of this paper is to develop a data mining process to evaluate non-trivial geologic settings for geosteering reservoir well placement. The well logs’ data was collected from multiple wells in a Norwegian North Sea field, where the reservoir rocks are characterized with high heterogeneities. Principal component analysis was used to recognize data pattern and extract underlying features. The extracted features are then into distinct groups using Hierarchical clustering (HC) analysis. A classification model, that is based on the deviance analysis, was constructed to build a criterion to identify each cluster within a set of well log data. The results show that the data mining approach sufficiently identified highly heterogeneous formations and can be used for geosteering applications. Classification trees defined quantitative decision criterion for the identified clusters. The approach is capable of distinguishing between potential and non-potential steering clusters, as the identified clusters have distinct decision criteria and effectively explain the variations within a section, as verified with the lithology described from core analysis.