三角洲前缘储层具有丰富油气资源，但其非均质性严重，已有建模方法难以准确建立三角洲前缘储层精细地质模型。以胜利油田典型分流砂坝型三角洲前缘储层为对象，开展多点地质统计建模方法研究。通过现代沉积、古代露头、水槽实验、卫星调查等资料建立分流砂坝型三角洲前缘储层地质知识库，确定不同储层结构形态数学函数和几何学参数分布规律。采用层次建模思路，设计分流砂坝型三角洲前缘储层训练图像自动建立方法，形成训练图像库，解决训练图像建立难题；通过理论和实际条件数据，考虑三角州前缘不同位置处沉积模式（数据事件）差异，从数据事件重复率以及其相对位置空间分布两个方面完成训练图像评价。最后，以计算机图形学理论为依据，以距离函数为工具，设计基于位置约束的函数，对训练图像进行尺度相对变换，解决建模中训练图像与模拟区域规模一致的限制。实现非平稳三角洲前缘储层多点地质统计预测。完善和发展多点地质统计建模理论和方法，服务油田生产。

Delta front deposits, with very complex components and serious heterogeneity, have accumulated rich oil and gas resources. The construction of reservoir geological model is very difficult and challenging. This project ,taking distributary bar type delta front reservoir in Shengli oilfield as research object, develops multiple point geostatistics (MPS)modeling method to solve the problem. The modern sedimentary, ancient outcrop, water tank laboratory and satellite survey data are collected to establish distributary bar type delta front reservoir geological knowledge base, and the geometric parameters and mathematical functions of shapes of different reservoir architectural elements are studied. Using hierarchical modeling ideas, automatic establishment method of training image are designed to form a training image database for MPS. Through theoretical and actual condition data, the repetition and relative position of data events are considered to evaluate the efficiency of training image . Finally, using computer graphics theory and taking the distance function as a tool, the location-based function are designed to constrain the selection of data event in training image during MPS modeling. Relative size transformation solves the limitation of the same size between the training images and modeling area ,and so the non-stationary reservoir prediction achieved in the practical study area. The study enlarge the modeling theory and provide a new designed MPS method for reservoir production.