Abstract: Aimed at inadaptability in current automatic identification models and algorithms of sedimentary microfacies,a new method is proposed by combination of genetic-BP algorithm and image process technology.Facies style is determined according to data of sampling well and interpretation results of experts.Minimal decision-making rule of rough set is used to choose pattern character and typical stylebooks.All kinds of standard model bases of sedimentary facies are built.Logging curves and stratum parameters are changed to image pattern by image process technology.Neural networks are introduced to extract and remember pattern character of curves automatically. Multi-layer forward neural networks are trained by combing BP and genetic algorithm.The gained network is of steadiness,fast study convergence speed, strong memory and generalization ability.Considered the multi-resolution of cambric facies,identification results nearest well and the same layers are consulted according to the result of small layer contrast.Fuzzy logic method is introduced to affirm and modify facies style according to the stratum rules in big circumstance of this region.The consistency of plane sedimentary facies and single well facies of small layer are assured.The model has a good adaptability to the problem of automatic identification of sedimentary facies. Test data of 15 wells from Daqing Oil Field show good results of the method.

Key words: sedimentary microfacies, automatic identification, neural networks, image process, BP algorithm, genetic algorithm, logging curve

摘要： 针对目前沉积微相自动识别模型和算法存在的某些不适应性,提出了一种基于遗传-BP算法与图像处理技术相结合的方法.根据取心井分析资料和专家解释结果确定区块微相类型,采用最小决策规则对模式特征指标和典型样本进行筛选,建立各类沉积微相标准模式库.利用图像处理技术将测井曲线和地质参数转化为图像模式,由神经网络自动提取和记忆曲线所表征的小层模式特征.用遗传和BP算法相结合的方法训练多层前馈神经网络,所得的神经网络稳定,学习收敛速度快,同时有很强的记忆能力和推广能力.对于过渡性微相在识别中存在的多解性,在小层对比基础上,参照邻井同层微相识别结果,在大环境下依据区块地质规律采用模糊逻辑推理方法确认和修正微相识别类型,保证平面沉积相和小层单井相的一致性.此模型对解决沉积微相自动识别问题具有良好的适应性.对大庆萨北油田15口井进行了资料处理,取得了较好的效果.

关键词: 沉积微相, 自动识别, 神经网络, 图像处理, BP算法, 遗传算法, 测井曲线