Improved well logs clustering algorithm for shale gas identification and formation evaluation

Abstract The identification of lithology, fluid types, and total organic carbon content are great priority in the exploration unconventional hydrocarbons. As a new alternative, further developed K-means type clustering method is suggested for evaluation shale gas formations. traditional approach cluster analysis mainly based on use Euclidean distance grouping objects multivariate observations into different clusters. high sensitivity L 2 norm applied to non-Gaussian distributed measurement noises well-known, which can be reduced by selecting more suitable as metrics. To suppress harmful effect non-systematic errors outlying data, Most Frequent Value robust statistical estimator combined with algorithm. Cauchy-Steiner weights calculated procedure measure weighted between objects, improves performance compared norm. At same time, centroids also average (using method), instead applying arithmetic mean. tested using synthetic datasets well observed wireline logs, mud-logging data core samples collected from Barnett Shale Formation, USA. experiment extremely noisy logs demonstrates that newly able separate geological-lithological units hydrocarbon formations provide additional information standard log analysis. It shown affected less outliers, allows efficient processing poor-quality an improved reservoirs.