神经网络方法确定玛北油田岩性油藏含油边界

用常规方法确定岩性油藏含油边界需要一定的井孔资料，导致勘探成本较高，为此引入神经网络方法。先对地震资料进行解释，提取地震特性参数，与已知井的试油结果一起组成神经网络的学习样本集，经训练并绘制测线剖面或平面图，最终确定出含油边界。应用该方法对准噶尔盆地玛北油田二叠系乌尔禾组和三叠系百口泉组油藏进行了含油边界的确定，结果表明，用该方法可以直观地在剖面图上确定油藏的含油边界，这样确定的含油边界能够满足储量计算的要求，且该方法较传统方法有很大的优越性。

Determining oil boundary of lithologic reservoir by neural network method in Mabei Oilfield

Certain borehole data are needed to determine the oil boundary of lithologic reservoir by conventional approaches, which leads to high exploration cost. The neural network method was introduced into this work. First, seismic data are interpreted and seismic characteristic parameters are picked up. Then, study samples of neural network are built by results of seismic interpretation and testing of oil. Finally, a survey line section or ichnograph is mapped out, and the oil boundary is determined. This method has been successfully applied to Wuerhe and Baikouquan reservoirs in Mabei Oilfield, Junggar Basin. The results show that the oil boundary can be determined visually on cross section by neural network method, which can fulfill the requirements of reserve calculation. The new method has more advantages than conventional approaches.