基于单分类支持向量机的优势储层评价方法

Favorable reservoir evaluation method based on one class support vector machine

周雪 张珊珊 李禄胜 李玉蓉 柳伟明 赵晓亮 董鹏

ZHOU Xue ZHANG Shanshan LI Lusheng LI Yurong LIU Weiming ZHAO Xiaoliang DONG Peng

延长油田股份有限公司志丹采油厂 中国石油大学(北京)油气资源与探测国家重点实验室

Zhidan Oil Production Plant, Yanchang Oilfield Co., Ltd., Yanan 716000, Shaanxi, China State Key Laboratory of Petroleum Resources and Prospecting, China University of Petroleum (Beijing), Beijing 102249, China

实际油藏非均质性强，高产井样本数量少而数据价值高。为充分利用高产井有效信息，使用单分类支持向量机(OCSVM)对高产井地质模型中的地质特征进行无监督学习，通过优化模型超参数，获得最优决策函数，基于该函数确定非均质油藏可能的高产区域分布进而来确定优势储层，从而为油藏开发及井位部署提供指导。案例研究结果表明，在井样本较少的情况下，OCSVM决策函数值与产量的相关性系数较高，使用OCSVM计算的决策值可以有效地确定全区高产区域分布并依据该分布确定优势储层。

Real oil reservoirs are of strong heterogeneity, there are few samples of high-yield wells and their data are of high value. In order to make full use of effective information of high-yield wells, this paper carried out unsupervised learning on the geological characteristics in the geological model of high-yield well by virtue of one class support vector machine (OCSVM). Then, the optimal decision function was obtained by optimizing the super parameters of the model. Based on this function, the distribution of possible high-yield zones of heterogeneous oil reservoirs was determined and accordingly the favorable reservoirs were determined, so as to provide guidance for oil reservoir development and well deployment. Case study results indicate that the correlation coefficient between OCSVM decision function value and production rate is higher, so the decision value calculated by means of OSCVM can effectively determine the distribution of high-yield zones in the whole area and accordingly determine favorable reservoirs when there are fewer well samples.

10.13639/j.odpt.2021.04.014