基于多属性融合方法的窄河道砂体精细刻画及储层预测

Fine Characterization of Narrow Channel Sand Bodies and Reservoir Prediction Based on Multi-Attribute Fusion Method

摘要: 蓬莱镇组为典型的浅水三角洲前缘沉积，河道砂体为其优势相带。为解决蓬莱镇组砂体纵横向变化快，对比难度大，河道砂体分布复杂等难题，增产增效，亟需对河道砂体进行精细刻画并完成储层的定量预测。针对该区河道砂体的地质地震特点，本文以测井标定地震，地震反馈测井，井震互馈精确划分小层。根据不同厚度砂体的振幅与地震频率的关系，分频刻画河道砂体并将不同频率切片利用RGB融合技术进行融合，进一步刻画河道边界。采用基于神经网络的智能属性融合预测砂体厚度，提高储层定量预测的精度，在一定程度上改善了单一属性分析存在的多解性问题。结果显示：1) 与原始数据体相比，高频、低频振幅属性分别对薄层、厚层砂体有更好的识别效果，进行分频RGB融合后，亮色河道边界更加清晰。2) 与单一属性相比，基于神经网络的智能属性融合技术的定量预测可靠性得到提升，与砂体厚度的相关系数可达0.784，较单一属性提升约10%。

Abstract: The Penglaizhen Formation is a typical shallow-water delta front deposit, and the channel sand bodies are its dominant facies belts. In order to solve the difficult problems such as the rapid vertical and horizontal changes of sand bodies in the Penglaizhen Formation, the great difficulty in correlation, and the complex distribution of channel sand bodies, and to increase production and efficiency, it is urgent to conduct a detailed characterization of the channel sand bodies and complete the quantitative prediction of reservoirs. In view of the geological and seismic characteristics of the channel sand bodies in this area, this paper uses well logging to calibrate seismic data and seismic data to feed back to well logging, and realizes the accurate division of thin layers through the mutual feedback between wells and seismic data. According to the relationship between the amplitude and seismic frequency of sand bodies with different thicknesses, the channel sand bodies are characterized by frequency division, and the slices of different frequencies are fused by using the RGB fusion technology to further delineate the boundaries of channels. The intelligent attribute fusion method based on neural networks is adopted to predict sand body thickness, which improves the accuracy of quantitative reservoir prediction and mitigates the non-uniqueness problem inherent in single-attribute analysis to a certain extent. The results show that: 1) Compared with the original data volume, the high-frequency and low-frequency amplitude attributes have better recognition effects on thin-layer and thick-layer sand bodies respectively. After the frequency division RGB fusion, the boundaries of bright channels are clearer. 2) Compared with a single attribute, the quantitative prediction reliability of the intelligent attribute fusion technology based on neural network is significantly improved, and the correlation coefficient with the thickness of the sand body can reach 0.784, which is about 10% higher than that of the single attribute.