基于GA-ANFIS的油气储层地震预测方法及应用
|
摘要
基于GA-ANFIS理论,将遗传算法与模糊神经网络技术有机地相结合,构成一种新的油气储层地震非线性预测方法。这种新的预测方法在油气储层预测中,利用地震数据和测井数据之间的非线性映射关系建立训练样本,将GA算法与ANFIS网络中的学习算法相结合,构成混合算法来优化ANFIS网络的前提参数和结论参数,并在遗传算法中加入禁忌搜索算法,这种混合算法自始至终将各算法按一定概率比例进行,其概率自适应变化,加快了网络收敛速度和提高了网络性能,获得了良好的预测效果。在测井数据约束下,应用所提出的方法对碳酸岩盐储层和砂岩储层分别进行了平面预测和剖面预测,并按储层有效性指数进行了储层分级,这种分级反映了储层的有效性和含油气状况,提高了油气储层的实际预测效果,是对油气储层预测技术的一种新发展,开拓了油气储层预测发展技术。
Based on GA-ANFIS theory, a new nonlinear seismic prediction method of oil/gas reservoir is presented by integrating the genetic algorithms with ANFIS neural networks. In reservoir prediction, the nonlinear mapping relationship between seismic and logging data is used to generate training samples, a mixed algorithm formulated by integrating the GA algorithm with the learning algorithm of ANFIS is used to optimize the prerequisite parameters and conclusion parameters, and TS algorithm is integrated with genetic algorithm. Each algorithms in the mixed algorithm always run by probability, which change in a self-adaptation manner, accelerating convergence and improving the performance of the network. The prediction effects are good. The new method is applied to lateral and vertical prediction of carbonate and sandstone reservoirs and the reservoirs are classified according to their validity indexes. This classification can reflect the effectiveness and oil/gas-bearing properties of reservoirs and improve the actual prediction effects of oil/gas reservoirs. This new method is a new development of reservoir prediction technology.
Based on GA-ANFIS theory, a new nonlinear seismic prediction method of oil/gas reservoir is presented by integrating the genetic algorithms with ANFIS neural networks. In reservoir prediction, the nonlinear mapping relationship between seismic and logging data is used to generate training samples, a mixed algorithm formulated by integrating the GA algorithm with the learning algorithm of ANFIS is used to optimize the prerequisite parameters and conclusion parameters, and TS algorithm is integrated with genetic algorithm. Each algorithms in the mixed algorithm always run by probability, which change in a self-adaptation manner, accelerating convergence and improving the performance of the network. The prediction effects are good. The new method is applied to lateral and vertical prediction of carbonate and sandstone reservoirs and the reservoirs are classified according to their validity indexes. This classification can reflect the effectiveness and oil/gas-bearing properties of reservoirs and improve the actual prediction effects of oil/gas reservoirs. This new method is a new development of reservoir prediction technology.
引文
[1]李勇,等.高分辨率非线性反演方法及应用研究[J].天然气工业,2004,24(3):58-60.
[2]李正文,等.沉积盆地有效储集层综合识别技术[M].成都:四川科学技术出版社,2002.
[3]李敏强,等.遗传算法的基本理论与应用[M].北京:科学出版社,2003.
[4]戴葵,等译.神经网络设计[M].北京:机械工业出版社,2003.
[5]赵太银.ANFIS理论与油气储集层非线性评价方法研究[D].成都:成都理工大学,2002.
[6]KITANO H.Designing neural networks using genetic al-gorithms with graph generation system[J].Complex Sys-tems,1990(4):461-476.
[7]JANGJ S R.ANFIS:adaptive-network-based fuzzyinfer-ence system[J].IEEE Transaction on System,Man and Cyberidge,1993(23):665-685.
[8]ANDERSON A.An introduction to neural network[R].Cambridge.MA:MIT press,1995.
[2]李正文,等.沉积盆地有效储集层综合识别技术[M].成都:四川科学技术出版社,2002.
[3]李敏强,等.遗传算法的基本理论与应用[M].北京:科学出版社,2003.
[4]戴葵,等译.神经网络设计[M].北京:机械工业出版社,2003.
[5]赵太银.ANFIS理论与油气储集层非线性评价方法研究[D].成都:成都理工大学,2002.
[6]KITANO H.Designing neural networks using genetic al-gorithms with graph generation system[J].Complex Sys-tems,1990(4):461-476.
[7]JANGJ S R.ANFIS:adaptive-network-based fuzzyinfer-ence system[J].IEEE Transaction on System,Man and Cyberidge,1993(23):665-685.
[8]ANDERSON A.An introduction to neural network[R].Cambridge.MA:MIT press,1995.
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