储层测井特征属性反演方法
|
摘要
储层预测是隐蔽油气藏勘探的一项关键技术,由于储层和非储层的波阻抗常常重叠而导致单纯使用波阻抗进行准确的储层预测非常困难。利用测井资料,针对岩性油气藏的具体地质情况,在井点构建能反映储层物性空间变化的特征属性,并利用神经网络映射技术反演此特征属性数据体。在具体实施过程中,利用多元逐步回归方法和交互校验技术筛选敏感的地震属性,能有效去除冗余信息,确定最佳属性组合;采用卷积算子将地震资料的低频信息和测井特征属性的高频信息相融合,可以有效地拓宽频带,提高反演结果的分辨率;用概率神经网络建立测井特征属性和地震属性之间的非线性映射关系,可以外推反演得到高精度的测井特征属性数据体。对比研究表明,反演的结果具有比较高的可靠性。在实际应用中,这套技术取得了明显的效果。
The reservoir prediction is a key technique in subtle oil and gas reservoir exploration. It was very difficult to predict accurately reservoir only by wave impedance because of the overlaps of the wave impedances of reservoirs and non-reservoirs. By use of log data, in consideraion of the geological conditions of lithologic oil and gas reservoirs and under the conditions of reflecting the character attribute of the spacial variation of petrophysical property by well distribution, the data volume of the character attribute may be inversed by applying the nerve network mapping technique. In the process of implementing the method, the sensitive seismic attributes might be selected by use of multivariate stepwise regression method and interactive check-up technique, thus eliminating the redundant seismic messages so as to determine optimal attribute assmblage; through combining the low-frequency messages of seismic data with the high-frequency messages of log character attributes by applying convolution operator, the frequence band might be effectively widened to enhance the resolution factor of inversion results; and, on the basis of setting up the non-linear mapping relation between the log character attribute and the seismic attribute by use of the probabilistic neural network method, the high accurate data volume of log character attributes might be achieved through extrapolating the inversion results. A comparative research indicated that these inversion results were of relatively high reliability. An obvious effectiveness has been obtained by applying this set of technique on the spot.
The reservoir prediction is a key technique in subtle oil and gas reservoir exploration. It was very difficult to predict accurately reservoir only by wave impedance because of the overlaps of the wave impedances of reservoirs and non-reservoirs. By use of log data, in consideraion of the geological conditions of lithologic oil and gas reservoirs and under the conditions of reflecting the character attribute of the spacial variation of petrophysical property by well distribution, the data volume of the character attribute may be inversed by applying the nerve network mapping technique. In the process of implementing the method, the sensitive seismic attributes might be selected by use of multivariate stepwise regression method and interactive check-up technique, thus eliminating the redundant seismic messages so as to determine optimal attribute assmblage; through combining the low-frequency messages of seismic data with the high-frequency messages of log character attributes by applying convolution operator, the frequence band might be effectively widened to enhance the resolution factor of inversion results; and, on the basis of setting up the non-linear mapping relation between the log character attribute and the seismic attribute by use of the probabilistic neural network method, the high accurate data volume of log character attributes might be achieved through extrapolating the inversion results. A comparative research indicated that these inversion results were of relatively high reliability. An obvious effectiveness has been obtained by applying this set of technique on the spot.
引文
1 李丕龙,陈冬霞,庞雄奇.岩性油气藏成因机理研究现状及展望.油气地质与采收率,2002;9(5):1~3
2 陈冬霞,庞雄奇,翁庆萍等.岩性油藏三元成因模式及初步应用.石油与天然气地质,2003;24(3):228~232
3 HampsonD P,SchuelkeJ S,QuireinJ A.Use of multiat tribute transforms to predictLog properties from seismicdata.Geophysics,2001;66(1):220—236
4 DraperN R,SmithH.Applied regression analysis.JohnWiley&SonsInc,1966
5 SpechtDonald.Probabilistic neural networks.NeuralNet works,1990;(3):109—118
6 SpechtDonald.A general regression neural network.IEEETransactions onNeuralNetworks,1991;2(6):568—576
7 徐旺林,魏立花等.沿层属性分析在储层研究中的应用.石油地球物理勘探,2002;37(增刊):83~87
2 陈冬霞,庞雄奇,翁庆萍等.岩性油藏三元成因模式及初步应用.石油与天然气地质,2003;24(3):228~232
3 HampsonD P,SchuelkeJ S,QuireinJ A.Use of multiat tribute transforms to predictLog properties from seismicdata.Geophysics,2001;66(1):220—236
4 DraperN R,SmithH.Applied regression analysis.JohnWiley&SonsInc,1966
5 SpechtDonald.Probabilistic neural networks.NeuralNet works,1990;(3):109—118
6 SpechtDonald.A general regression neural network.IEEETransactions onNeuralNetworks,1991;2(6):568—576
7 徐旺林,魏立花等.沿层属性分析在储层研究中的应用.石油地球物理勘探,2002;37(增刊):83~87
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心