摘要
渤海蓬莱PL油田勘探程度低,可用井资料有限,采用常规阻抗反演方法预测油气效果不理想。此文尝试用分频多属性联合神经网络算法反演进行油气预测,效果较明显。首先利用频谱分解技术将地震信号从时间域转换到频率域,寻找到油气敏感的高中低频率组合;再将多个与油气异常相关的属性进行相关性分析,优选出效果最好的一组属性组合,通过BP神经网络算法建立起地震属性与含油饱和度曲线之间的非线性相关性,与钻井证实的含油层段对比验证方法的可靠性;最后用建立起来的非线性关系对目的层段油气检测,预测出含油气潜力砂体,指导井位部署。该方法在地震资料品质较好,砂地比适中的浅层含油气层段具有普遍适用性。
The well data available from Penglai Oilfield of Bohai basin is limited due to the low degree of exploration.Therefore,the effect of conventional impedance inversion used to predict the oil and gas is not ideal.The authors attempted to use the combination of frequency division multi-attribute and neural network algorithm for the prediction of oil and gas,and achieved good effect.Firstly,the seismic signals were transformed from time domain to frequency domain by the spectrum decomposition technique,and the combination of high,medium and low frequency sensitive to the hydrocarbon was determined.Then,analyzed the correlation of attributes that were associated with hydrocarbon anomalies,optimized the best combination of attributes,established the nonlinear correlation of seismic attributes and oil saturation curve by BP neural network algorithm,and verified the effectiveness of this method by well data.Finally,the nonlinear correlation was used to detect the hydrocarbon,predict the sand bodies with potentials and guide the deployment of wells.The method has good applicability in the shallow hydrocarbon-bearing strata with moderate sand ratio and good quality of seismic data.
作者
马佳国
李才
王腾
赵志平
蒋志恒
李博
MA Jiaguo;LI Cai;WANG Teng;ZHAO Zhiping;JIANG Zhiheng;LI Bo(Tianjin Branch of CNOOC Ltd.,Tianjin 300452,China)
出处
《海洋石油》
CAS
2018年第3期13-17,26,共6页
Offshore Oil
关键词
BP神经网络
烃类检测
频谱分解
多属性组合
岩性油气藏
BP neural network
hydrocarbon detection
multi-attribute combination
lithological reservoir
