基于BP神经网络的油藏流场评价体系研究

Study on reservoir flow field evaluation system based on BP neural network

 姜瑞忠¹ 王平¹ 候玉培² 梁宇³

中国石油大学（华东）石油工程学院，山东 青岛 266580 中国石化胜利油田分公司地质科学研究院，山东 东营 257051 怀俄明大学，美国 怀俄明 WY82071

室内试验及矿场测试均显示，注水开发油田储层参数会随注水冲刷而发生变化，储层孔隙度及渗透率均会增大,形成优势流场，造成注入水的无效循环，降低开发效果。准确识别并评价优势流场，对于有针对性地采取高含水期剩余油挖潜措施具有重要意义。基于BP神经网络技术的方法对优势流场进行评价,各指标的权重值由BP神经网络经样本训练后获得,不再由层次分析法或专家打分法确定,消除了主观因素的影响，客观反映了各评价指标对优势流场的影响。将该评价体系应用于埕东油田试验区进行优势流场分级及评价，针对评价情况部署了油田剩余油挖潜方案,经数值模拟证明，剩余油挖潜效果较好。

Both experiments and pilot tests show that the reservoir parameters will change with the long?鄄term water injection process in waterflooding oilfields. Reservoir porosity and permeability will increase, dominant flow field will be formed, the invalid circulation of water will occur and development effect will be reduced. It is of great importance to accurately indentify and classify dominant flow field for further tapping the potential of remaining oil in high water cut stage. BP neural network is used to evaluate dominant flow field, the weight of each factor that affects dominant flow field is obtained through BP neural network samples training instead of AHP and expert scoring method, which is more objective and can reflect objectively the influence of each factor on dominant flow field. Using this method in Chengdong Oilfield, dominant flow field is evaluated and classified, targeting EOR methods. The numerical simulation results show that the effects of tapping the remaining oil potential are good.

10.6056/dkyqt201203011