神经网络在低渗透油田试井解释中的应用

Application of neural network in well test analysis in low permeability oilfield

王安辉 宇淑颖 张英魁 王龙源4苗德顺 盛国军 刘家君 王琳芳

Wang Anhui Yu Shuying Zhang Yingkui Wang Longyuan Miao Deshun Sheng Guojun Liu Jiajun Wang Linfang

A油田是吉林油区开发较好的典型低渗透砂岩油藏,其试井解释比较复杂,压力恢复曲线出现径向流的井次仅占总井次的20%～30%。图形识别+神经网络BP算法+试井解释软件三位一体的联合技术能使未出现径向流的大部分井的压力恢复资料得到很好应用。该技术具体步骤为:(1)分析解释有径向流的井的双对数图和半对数图,找出续流段的伪斜率(m)、拐点处的伪斜率(m)、过渡段的伪斜率(m)和径向流直线段斜率(m);(2)利用神经网络BP算法,构建m,m,m与m之间的数学关系;(3)将未出现径向流的井的基础测试资料录入到试井解释软件中,求出m,m,和m,利用BP算法求出m;(4)把以上参数代入进行拟合,直到双对数图、半对数图和历史拟合图三条曲线完全拟合为止。

"A" oilfield is a typical low permeability sandstone oil reservoir that has relatively successfully been developed in Jilin oilfield area.The well test interpretation is relatively complicated.The times of radial flow occurring on wells'pressure build-up curves account for only 20%～30% of the total times occurring in all tested wells.This paper introduces an integrated interpretation technology by integrating pattern recognition,neural network BP algorithm and well test interpretation software.It can specifically divided into the following steps:(1)analyze and interprete the bilogarithmic and semilogarithmic diagrams of wells with radial flows,and find out the pseudoslopes(m,m,m and m)in the continuous flow section,at the flex point,in the transitional section and on the straight line section of radial flow;(2)applying the neural network BP algorithm to construct the mathematical relation among m,m,m and m;(3)input the basic testing data of the wells without radial flow into the well test interpretation software,derive the m,m and m,and then derive m with BP algorithm;(4)substitute the parameters mentioned above and fitting them through to the three curves in the bilogarithmic and semilogarithmic diagrams,and in historic fitting diagram to be fitting to one another.

10.11743/ogg20040320