Incorporation of Spatial Characters into Volcanic Facies and Favorable Reservoir Prediction

Abstract

Abstract Compared with clastic reservoirs, volcanic reservoirs are of higher heterogeneity. Lithological facies is one of most important indicators of favorable volcanic reservoir. Traditionally, facies are identified by core observation or log data classification. However, the spatial distribution characters and geological priors, which are important in early stage of exploration, are seldom quantitatively incorporated in facies prediction. Based on previous work, a new methodology is developed to incorporate volcanic spatial information with few well log data (3 wells) to improve the facies prediction accuracy. Such method is applied to volcanic clastic reservoir of Yingchen member, Cretaceous, Xinshan fault depression, northeastern China. Based on the assumption that the residual thickness of volcanic formation is controlled by eruption center, indicator simulation of seismic thickness data is used to estimate the prior probabilities of occurrence of facies. Such prior is combined with well log facies data within a Bayesian framework. Artificial neural network, beta-Bayesian method as well as discriminant analysis algorithm are used to predict log-based facies. Confidence analysis is applied to evaluate the facies prediction. Analysis of variance (ANOVA) proves the overall prediction accuracy is above 82%. To incorporate updated facies data with geological information (diagenetic facies, potential fractural area and volcanic residual thickness), Indicator kriging algorithm is applied to generate reservoir quality index. Three favorable reservoirs are predicted based on above methodology, which is proved by the subsequent drilling.