基于计算智能的油水识别技术的研究
|
摘要
计算智能即用计算机模拟和再现人类的某些智能行为,主要包括神经网络、模糊技术和进化计算等技术.针对地震勘探中油水识别误差较大的问题,提出一种基于计算智能的神经网络识别技术,首先进行地震波属性提取,然后对所取属性进行预处理,再用神经网络进行学习训练,最后进行油水识别或预测.实际试验选取我国油田某区的典型地震勘探数据作为试验资料,提取出六种时间、振幅等有用属性,采用BP和LM两种算法的进行网络训练比较,得到LM优于BP算法.实际应用表明,基于计算智能的此种油水识别方法是切实可行的,可以达到识别系统理想的精度要求,而且起到节省成本、提高效率等功效,在油水识别方面效果显著.
Computational Intelligence (CI), including Fuzzy, Neural Networks (NN), and Evolutionary Computation (EC), is to simulate some human intelligent behavior by using computer. In the seismic prospecting, there is still big error in oil-water recognition at present. So an intelligent recognition method is presented based on computational intelligence, including the following steps: first, seismic wave feature extraction; second, preprocessing; third, neural network training; forth, the oil-water recognition or forecast. Six features based on time and amplitude are extracted. Both BP and LM neural network are built to give an exact result and have a compare. The LM algorithm is super to that of BP. Actual ap-plication examples in an oil field show that the oil-water recognition method proposed based on the computational intelli-gence is practical. It can reach the ideal precision request of the recognition system. Meanwhile, it helps to save cost and improve efficiency. The method is significantly valuable in the oil-water recognition field.
Computational Intelligence (CI), including Fuzzy, Neural Networks (NN), and Evolutionary Computation (EC), is to simulate some human intelligent behavior by using computer. In the seismic prospecting, there is still big error in oil-water recognition at present. So an intelligent recognition method is presented based on computational intelligence, including the following steps: first, seismic wave feature extraction; second, preprocessing; third, neural network training; forth, the oil-water recognition or forecast. Six features based on time and amplitude are extracted. Both BP and LM neural network are built to give an exact result and have a compare. The LM algorithm is super to that of BP. Actual ap-plication examples in an oil field show that the oil-water recognition method proposed based on the computational intelli-gence is practical. It can reach the ideal precision request of the recognition system. Meanwhile, it helps to save cost and improve efficiency. The method is significantly valuable in the oil-water recognition field.
引文
[1]Skowron A,winiarski R S.Information Granulation and pattern Recognition[A].K Pal Sankar,L Polkowski,A Skowron(eds.),Rough Neurocomputing:Techniques for Computing with Words,Cognitive Technologies[M].Springer-Verlag,2003,43-636.
[2]夏克文,沈钧毅,李昌彪.一种基于计算智能的油水层识别方法[J].计算机工程与应用,2005,41(5):190-192.
[3]朱广生.地震资料储层预测方法[M].北京:石油工业出版社,1995.
[4]夏克文,沈钧毅,李昌彪.样本信息中的属性约简方法研究[J].西安交通大学学报,2005,39(6):558-602.
[5]靳蕃.神经计算智能基础[M].成都:西南交通大学出版社,2000.
[6]夏克文,沈钧毅,李昌彪.前向神经网络隐含层节点数的最优算法[J].计算机科学,2005,32(10):143-145.
[7]Huang S C,Huang Y F.Bounds on number of hidden neurons in multiplayer perceptrons[J].IEEE Trans on Neural Networks,1991,2(1):47-55.
[2]夏克文,沈钧毅,李昌彪.一种基于计算智能的油水层识别方法[J].计算机工程与应用,2005,41(5):190-192.
[3]朱广生.地震资料储层预测方法[M].北京:石油工业出版社,1995.
[4]夏克文,沈钧毅,李昌彪.样本信息中的属性约简方法研究[J].西安交通大学学报,2005,39(6):558-602.
[5]靳蕃.神经计算智能基础[M].成都:西南交通大学出版社,2000.
[6]夏克文,沈钧毅,李昌彪.前向神经网络隐含层节点数的最优算法[J].计算机科学,2005,32(10):143-145.
[7]Huang S C,Huang Y F.Bounds on number of hidden neurons in multiplayer perceptrons[J].IEEE Trans on Neural Networks,1991,2(1):47-55.
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