Abstract: The channel prediction of Fuyu reservoir has been always highly emphasized. Using the conventional seismic attribute to predict channel is difficult to achieve the precision requirements because of the complex fault and the fast lithofacies phase in Fuyu reservoir. Aiming at this weak problem, the method of combing gray correlation analysis and support vector machines is used to establish a set of technical process which is suitable for the prediction of fluvial reservoirs under complicated geological conditions. In Fuyu test area in X reservoir of Daqing oilfield as an example, firstly, conventional seismic attribute of sedimentary unit dimensionless, obtained by the method of gray correlation analysis of the seismic attribute correlation factor, the greater the degree of correlation, indicating that the response probability of attribute river is higher. On this basis, the optimal correlation factor sequence a accumulation of large properties, first order accumulative sequence is generated, used as the input into the support vector machine training sample, so as to complete construction of support vector machine river forecast model. Drilling confirms that the prediction based on Gray Correlation Support Vector Machine has a larger coincidence rate of drilling. Combined with the superiority of seismic inversion to predict the channel sand boundary, supplemented by data of core, well logging and mud logging data to complete sedimentary microfacies in X Test Area Fuyu reservoir. Meanwhile, drilling further confirms the reliability of predicting channel and then the industrial oil flow well is successfully obtained. The results of comprehensive research show that this method is suitable for high channel prediction accuracy. It can be used as a better channel prediction method under complicated geological conditions.

Key words: Grey relational analysis, Support vector machine, Channel, Fuyu oil layer, Daqing oilfield

摘要： 扶余油层河道预测一直被高度重视，由于扶余油层断裂复杂、岩相相变快等因素影响，常规地震属性预测河道难以达到精度的要求。针对这一薄弱的问题，利用灰色关联分析与支持向量机结合的方法，建立一套适用于复杂地质条件的河流相储层预测技术流程。以大庆油田X试验区扶余油层为例,首先对沉积单元的常规地震属性进行量纲化处理，运用灰色关联分析方法得到的各地震属性关联主因子，关联度越大，说明响应河道的属性概率就越高。在此基础上，对优选关联因子较大的属性数列进行一次累加，生成一阶累加序列，用以作为支持向量机的输入训练样本，从而完成支持向量机河道预测模型的构建。经钻井证实，基于灰色关联支持向量机预测的钻井符合率较高，并结合地震反演预测河道砂边界的优势，辅以岩心、测井、录井等资料，从而完成X试验区扶余油层沉积微相的实现。同时,后钻井进一步证实了预测河道的可靠性，成功获得工业油流井。综合研究表明，此方法用于河道预测精度较高，可作为复杂地质条件下一种较好的河道预测方法。

关键词: 灰色关联分析, 支持向量机, 河道, 扶余油层, 大庆油田