基于概率神经网络的烃源岩TOC预测——以珠江口盆地陆丰南区为例

Prediction of total organic carbon in source rocks by probabilistic neural network: a case study of southern Lufeng area in Pearl River Mouth Basin

石创 朱俊章 龙祖烈 秦成岗 史玉玲 黄玉平

中海石油（中国）有限公司深圳分公司，广东 深圳 518054）

Shenzhen Branch of CNOOC Ltd., Shenzhen 518054, China)

烃源岩评价在油气成藏和资源潜力研究中起着关键性作用，而总有机碳质量分数（TOC）是烃源岩评价的基础和影响油气资源评价的关键参数。海上油气勘探因钻井和取样数量的限制，难以获得连续的TOC数据。通过地化￣测井￣地震联合优选地震属性参数，基于概率神经网络（PNN）对陆丰南区烃源岩TOC进行了预测。结果表明：文昌组下段发育优质烃源岩，其中文四段烃源岩是陆丰凹陷南部油气资源的主要贡献者。运用PNN神经网络预测烃源岩TOC，可获得高精度烃源岩TOC三维数据体，充分揭示了烃源岩有机质丰度的非均质特点，有效弥补了海上少井区烃源岩评价的缺陷，为精细油气资源潜力评价提供了一种新的尝试。

he evaluation of source rocks plays a key role in the research of hydrocarbon accumulation and resource potential; however, the total organic carbon, the basis of source rock evaluation, is the key parameter to hydrocarbon resource evaluation. Due to the limitation of drilling and sampling quantity, it is difficult to obtain continuous TOC for offshore petroleum exploration. The seismic attribute parameters were optimized by comprehensively applying geochemistry, logging and geophysics, then TOC prediction of source rocks in southern Lufeng area was finished by probabilistic neural network. The result shows that high quality source rocks are developed in Lower Wenchang Formation, of which the source rocks of the fourth Member are the main contributors of oil and gas resources in the southern Lufeng area. Using probabilistic neural network to predict TOC of source rocks can obtain high?鄄precision TOC volume of source rocks, fully reveal the characteristics of organic matter abundance heterogeneity of source rocks, effectively eliminate the shortcomings of offshore source rock evaluation, and provide a new attempt for fine oil and gas resources potential evaluation.

10.6056/dkyqt201905004