利用神经网络技术预测剩余油分布

在油田开发过程中,剩余油饱和度是一个变量,且受多种因素的共同控制,预测难度较大。针对构造背景相对简单、分层测试资料比较齐全的海上非注水开发的典型边、底水油藏,初步研究出了一套根据主要的多种动、静态参数并应用神经网络技术预测含水饱和度随时间变化速率的方法,进而根据含水饱和度变化率来预测不同时期的剩余油饱和度分布。在我国海上某油田进行了应用实例验证,取得了比较满意的效果。

Forecast of remaining oil distribution by using neural network technology

During the development of oilfield, the remaining oil saturation is a variable dominated by many kinds of subjective and objective factors, so it is difficult to be forecasted. For the classical edge and basal water reservoirs in non-water flooding with relatively simple structure and abundant stratified test data, a new method for forecasting the change rate of water saturation by using dynamic and static parameters as well as artificial neural network technology is proposed. The distribution of remaining oil saturation in different stage can be forecasted according to the change rate of water saturation. An applied example in a offshore oilfield proved the reliability of this method.