地震多属性融合技术在贝尔凹陷储层预测中的应用

地震属性分析是一种快速、有效的储层预测方法，但地震单属性储层预测多解性强，尤其对多物源、相变快、多期次火山活动、岩石成分较复杂的断陷盆地，预测精度会明显降低。提出了基于沉积特征分区域的地震多属性融合储层预测方法。首先利用地震波形分类技术将研究区按沉积特征划分为不同区域，然后对不同区域分别统计地震多属性与储层的相关系数并进行线性拟合，最终将各区域拟合结果综合为全区储层预测结果。在海拉尔盆地贝尔凹陷南屯组储层预测中．利用164口井目的层砂地比数据及优选的7种地震属性，采用逐步线性回归法进行全区拟合，其相关系数仅为0．52；采用基于沉积特征分区域地震多属性融合技术进行储层定量预测，并将其综合为全区储层预测结果，其相关系数达0．85，且储层预测结果与钻遏储层发育情况及沉积规律吻合较好，说明该方法可明显提高储层预测精度。

Application of seismic multi-attribute fusion technique to the reservoir prediction in Beier Depression

Seismic attribute analysis is an efficient method for reservoir prediction. However, using only one attribute to predict reservoir leads to multiple-understanding, especially for the fault-depressed basin which has the charac- teristics of multi-provenances, fast variation of sedimentary facies, multi-period volcanic movements and complicated lithologies. All of these factors result in poor prediction accuracy. Therefore, this paper proposed a seismic multi-attribute fusion technique based on different sedimentary characteristics and separated regions for reservoir prediction. Firstly, according to the sedimentary characteristics, different regions were divided with seismic waveform classification technique. Then, the relation statistics between seismic multi-attribute and reservoir information were carried out and the linear fitting was completed for the further step. At last, the separated fitting results were integrated into the final prediction results for the whole area. For the reservoir prediction of Nantun Formation in Beier Depression, Hailaer Basin, stepwise linear regression was put into use on the basis of sand data from 164 wells and the selected 8 seismic attributes. Its correlation coefficient is only 0.42. With the seismic multi-attribute fusion technique based on sedimentary characteristics and selected regions, the whole area fitting was carried out. Its correlation coefficient reaches 0.89 and the reservoir prediction result fits well with the reservoir development and the sedimentary rules in drilled wells, which indicates that the method can improve the accuracy of reservoir prediction.