代数神经网络在岩性识别中的应用
|
摘要
针对辽河油田L区块砂砾岩储层岩性复杂多变、纵向非均质性严重的特点,采用代数神经网络算法,对该区块的岩性进行识别。代数神经网络算法通过优选激励函数和应用代数算法计算权值,将复杂的非线性优化问题转化为简单的代数方程组求解问题,从而克服了BP神经网络易陷于局部极小、收敛速度慢和耗费时间长的问题,提高了计算的精度和速度。并且利用训练好的代数神经网络对预测样本进行预测,识别的准确率达到92.0%。结果表明,代数神经网络可以很好的用于辽河油田L区块砂砾岩储层的岩性识别。
Based on the characteristics of complicated lithology and strong vertical heterogeneity of sand-conglomerate reservoir in L area of Liaohe oilfield,the algorithm of algebra neural networks is used to identify the lithology.The weight of algebra neural networks is calculated by preferred exciter functions and algebraic algorithm,and the algorithm transforms the complicated nonlinear optimization problem into linear algebraic equations and the adaption speed and accuracy are improved,and the algorithm overcome the problems that BP neural networks may easily fail into local minimum point and converge slowly and cost long time.The trained algebra neural networks are used to predict samples,and the identification accuracy rate reaches 92%.The result shows that algebra neural networks can be well used to identify the lithology in sand-conglomerate reservoir of L area in Liaohe oilfield.
Based on the characteristics of complicated lithology and strong vertical heterogeneity of sand-conglomerate reservoir in L area of Liaohe oilfield,the algorithm of algebra neural networks is used to identify the lithology.The weight of algebra neural networks is calculated by preferred exciter functions and algebraic algorithm,and the algorithm transforms the complicated nonlinear optimization problem into linear algebraic equations and the adaption speed and accuracy are improved,and the algorithm overcome the problems that BP neural networks may easily fail into local minimum point and converge slowly and cost long time.The trained algebra neural networks are used to predict samples,and the identification accuracy rate reaches 92%.The result shows that algebra neural networks can be well used to identify the lithology in sand-conglomerate reservoir of L area in Liaohe oilfield.
引文
[1]范宜仁、黄隆基、代诗华,交会图技术在火山岩岩性与裂缝识别中的应用[J],测井技术,1999,23(1),53-56.
[2]Hamada,G.M.,AI-Awad,M.N.J.,and Almalik,M.S.Log evaluation of low-resistivity sandstone reservoirs[J],SPE70040,2001.
[3]张晓峰、范晓敏,交会图和Carbon软件在火山碎屑岩识别中的应用[J],吉林大学学报(地球科学版),2007,37(增刊).
[4]张丽华、潘保芝、刘思慧等,梨树断陷东南斜坡带砂砾岩岩性识别方法研究[J],测井技术,2012,36(4),370-372.
[5]李伟才、姚光庆、黄银涛等,文昌13-1油田低阻油层测井岩性识别方法研究[J],石油天然气学报,2012,34(12).
[6]杨立强、王志章、熊琦华,复杂岩性储层参数评价中神经网络技术的应用[J],地球物理学进展,2003,18(1),44-48.
[7]王祝文、刘菁华、任莉,基于K均值动态聚类分析的地球物理测井岩性分类方法[J],东华理工大学学报(自然科学版),2009,32(2),152-156.
[8]李盼池、王海英、杨雨,基于量子神经网络的油田水淹层识别方法[J],计算机应用与软件,2012,29(5):41-43.
[9]杨辉、黄健全、胡雪涛等,BP神经网络在致密砂岩气藏岩性识别中的应用[J],油气地球物理,2013,11(1),39-42.
[10]张代远,神经网络新理论与算法,清华大学出版社,2006.
[11]赵忠伟、倪骁骅、孙美艳,改进代数算法神经网络在机器人逆运动学中的应用[J],机械设计与制造,2010,07,62-64.
[12]靳玉萍,基于代数神经网络的煤自然预测[J],数学的实践与认识,2013,43(18),122-128.
[13]龚菲、赵又群、范平等,基于实神经网络代数算法的汽车逆动力学仿真[J],农业装备与车辆工程,2013,51(12),1-4.
[2]Hamada,G.M.,AI-Awad,M.N.J.,and Almalik,M.S.Log evaluation of low-resistivity sandstone reservoirs[J],SPE70040,2001.
[3]张晓峰、范晓敏,交会图和Carbon软件在火山碎屑岩识别中的应用[J],吉林大学学报(地球科学版),2007,37(增刊).
[4]张丽华、潘保芝、刘思慧等,梨树断陷东南斜坡带砂砾岩岩性识别方法研究[J],测井技术,2012,36(4),370-372.
[5]李伟才、姚光庆、黄银涛等,文昌13-1油田低阻油层测井岩性识别方法研究[J],石油天然气学报,2012,34(12).
[6]杨立强、王志章、熊琦华,复杂岩性储层参数评价中神经网络技术的应用[J],地球物理学进展,2003,18(1),44-48.
[7]王祝文、刘菁华、任莉,基于K均值动态聚类分析的地球物理测井岩性分类方法[J],东华理工大学学报(自然科学版),2009,32(2),152-156.
[8]李盼池、王海英、杨雨,基于量子神经网络的油田水淹层识别方法[J],计算机应用与软件,2012,29(5):41-43.
[9]杨辉、黄健全、胡雪涛等,BP神经网络在致密砂岩气藏岩性识别中的应用[J],油气地球物理,2013,11(1),39-42.
[10]张代远,神经网络新理论与算法,清华大学出版社,2006.
[11]赵忠伟、倪骁骅、孙美艳,改进代数算法神经网络在机器人逆运动学中的应用[J],机械设计与制造,2010,07,62-64.
[12]靳玉萍,基于代数神经网络的煤自然预测[J],数学的实践与认识,2013,43(18),122-128.
[13]龚菲、赵又群、范平等,基于实神经网络代数算法的汽车逆动力学仿真[J],农业装备与车辆工程,2013,51(12),1-4.