Abstract: Deep-water turbidity channels are rich in oil and gas,and a hot area of hydrocarbon exploration and development.The deep-water turbidity channels in Angola are taken for reservoir modeling based on multi-point geostatistics.Alluvsim algorithm based on sedimentary process was improved to capture deep-water channels' sedimentation and migration characteristics.First,the integral migration and deposition processes of turbidity channels were simulated by adding integral migration algorithm to Alluvsim algorithm.Second,the curvature index constraint was also added to represent the original curvature characteristics of turbidity channels,to simulate the morphology of fewer crevasses and higher sinuosity for the channels of Angola deep waters.In addition,we obtained the morphology and statistical characteristic parameters of turbidity channels in the study area through literature surveys on the geological knowledge database of deep-water turbidity channels in Angola.The so acquired training images of turbidity channels in the study area improved Alluvsim algorithm.Thereafter,a three-dimensional geological model of the deep-water channels in Angola was established based on Snesim algorithm of multi-point geostatistics,to reproduce the spatial distribution of these channels,in which the training image was the model,logging data were the conditions,and seismic data were the trend constraint.The inspection of the model result shows that the variogram calculated by the geological model is in good agreement with that obtained by seismic attributes,and the geological model set up in this study may be used for subsequent reservoir engineering and numerical simulation research to guide oilfield exploration and development.Apart from providing a complete modeling workflow for the application of multi-point geostatistics to actual reservoir modeling,the study also demonstrates the technical feasibility for the automatic generation of turbidity channel training images.

Key words: variogram, three-dimensional training image, multi-point geostatistics, algorithm improvement, deep-water channel, depositional process