地质风险评价中的不确定性分析

Uncertainty evaluation of petroleum risk assessment

盛秀杰 金之钧 徐忠美 蒋瀚

Sheng Xiujie Jin Zhijun Xu Zhongmei Jiang Han

中国石油化工股份有限公司 石油勘探开发研究院, 北京 100083 2. 中国地质大学(北京) 信息工程学院, 北京 100083

SINOPEC Petroleum Exploration & Production Research Institute, Beijing 100083, China 2. Information Engineering Institute, China University of Geosciences (Beijing), Beijing 100083, China

地质风险概率法被广泛应用于油气资源评价中的地质风险评价。由于很难直接定义每一地质评价因子的评价(“打分”)概率模型，专家给出的都是每一评价因子的确定性评价值(“点估计”)—— 单一的、确定性评价结果并不能够很好地反映目标区域复杂地质情况的多种可能性，不可避免地出现含油气有无的高估或低估的情况。结合油气资源一体化评价软件(PetroV)的开发经验，从如何更客观地去描述地质评价因子的不确定性入手，阐述如何采用多种不确定性分析技术，更好地融合、体现专家经验以及客观表达地下复杂地质情况，从而提高地质风险概率法的不确定性表达能力：(1)通过自信度转换数学模型体现专家主观认知的不确定性；(2)将专家的定性认知进行合理的知识规则化转换，充分量化地质模型的不确定性；(3)基于地质风险概率法数学模型，利用蒙氏模拟计算符合概率分布特征的地质风险评价结果；(4)据地质风险不确定性评价结果，可获取目标区域地质风险评价的多分位评价值，为后续的勘探决策给出尽可能全面的决策方案。

The margin and condition probability analysis is broadly applied to the geological risk evaluation for an immature play with joining the "success" probabilities subjectively specified for those independently involved geological factors. Considering that it is difficult to specify a reasonable scoring distribution curve for each geological factor, this method contributes a "point" estimation about whether there exists petroleum resource. Obviously, subjected to lack efficient ways to encode the information about geological multi-scene of subsurface and possibi-lities for each geological factor, the above crisp estimation conclusion would be generally either higher or lower. In order to enhance the capability of uncertainty expression of geological risk evaluation, this paper presents three heuristic mathematic models to deeply quantify the understanding of a geological expert and objectively delineate the possibilities of subsurface occasion, etc. Meanwhile, the uncertainty assessment methods discussed also de-monstrate a reasonable uncertainty evaluation process from subjective guess to objective prediction in order to shrink the uncertainty of evaluation as more as possible. Firstly, the double linear conversion between subjective inference for each factor and quantized confidence shares an efficient alternative to describe subjective uncertainty. Next, the specification of multi-value model and setup of corresponding fuzzy rules for each geological factor may accurately and honestly reflect the worldly uncertainty of subsurface multi-scene while matching the domain expert’s understanding as more as possible. At last, Montecarlo method randomly joins the objective uncertainty distribution curve of each factor and shares experts with quantiles evaluation which would benefit incoming reasonable exploration solution. As a conclusion, this paper not only investigates how to make full scale uncertainty evaluation for geological evaluation, but also expands a new horizon about geological risk uncertainty research.

国家科技重大专项(2011ZX05005-001-004)资助。

https://doi.org/10.11781/sysydz201405633