A three-dimensional model of deep-water turbidity channel in Plutonio oilfield, Angola: From training image generation, optimization to multi-point geostatistical modelling

Abstract

The deep-water turbidity channel system in the Oligocene O73 sand layer in the Plutonio oilfield in Angola of west Africa has good oil and gas reservoir capability. As the well spacing is large due to operation cost, and the seismic resolution of the target layer is relatively low, identification of turbidity channel boundary was difficult. To establishes a fine model is a problem faced by the modelers. In this paper, a complete multi-point geostatistical modeling flowchart is presented for the construction of fine deep-water turbidity channel model. According to the investigation of seismic, well logging, drilling data and references in the study area, the genesis, reservoir scale, lithofacies type, morphological characteristics and evolution of the deep-water turbidity channel of the target layer were obtained. The improved Alluvsim algorithm was utilized to generate a set of adaptive deep-water turbidity channel training images with specified curvatures. Net-to-gross, variogram and a high-order statistics method based on the single data event were compared to evaluate and sort the training images. Then, the Snesim method was utilized to construct the models using the sorted compatible training images with a same set of parameters. At last, the quality of different models was compared based on the seismic attribute slices and a composite correlation coefficient optimization method based on seismic data. The results suggest that the improved Alluvsim algorithm can reproduce the channel morphology and stacking patterns better. Compared with the first-order and second-order statistics comparison method, the method using higher-order statistics of single data event taken spatial patterns into consideration and sorted training images better. The reservoir model was established using the optimal selected training image which has a high correlation with seismic attributes, and the model is more consistent with the characteristics of subsurface reservoirs. At the same time, this paper also provides a complete research program for the application of multi-point geostatistical modeling method which is from training image generation, optimal training image selection and multi-point geostatistics modeling to realizations selection for practical subsurface reservoir modelling.