面向模式识别的过程神经网络模型构建及在沉积微相判别中的应用

发布时间：2018-05-20 02:00

本文选题：模式识别 + 模糊推理　；参考：《东北石油大学》2017年硕士论文

【摘要】：大庆油田由于取心成本高,取心井数目少,因而通过分析取心样品来判别沉积微相的微相类型存在很大的局限性。本课题利用油田中已经取心的油井的测井曲线数据,结合专家知识,研究过程神经网络的模式识别方法及应用技术,将研究成果应用于非取心井的沉积微相判相中,实现对某一区块的石油储量及开采价值评估。结合沉积微相实际分析发现,就判相数据而言,除了测井曲线相关定量数据外,可能专家经验等定性信息同样会对判别结果产生影响,而且微相类型本身也具有模糊性和随机性,因此如果只采用定量识别的模型进行判相,那么识别结果的误差可能会较大。针对上述不足,本文题出了模糊推理过程神经网络模型和基于云变化的混合计算过程神经网络模型来实现对沉积微相的判别;一方面,针对有专家经验的评判规则数据,利用模糊推理可实现对测井相的定性信息进行定量处理以简化判别规则,提取有效的判别数据,进而提高沉积微相判别的精度;另一方面,针对有模糊信息的数据,利用云模型可将判相数据中的定性信息进行转化,从而纳入计算过程,保证了原始数据的完整性与客观性;考虑到测井相的数据是随深度变化的曲线这一特征,因而采用过程神经网络的过程式输入优势,借助其具有对时、空二维信息处理能力和不断优化过程神经网络的学习机制来提高沉积微相判别的准确度。课题以沉积微相数据要素为基础,以沉积微相识别问题所建立的模型和算法为核心,按照软件工程的开发模式,进行软件设计和集成开发后形成了沉积微相模式识系统,并取得了较好的实验分析结果。课题的研究对于沉积微相判别资料实际处理,微相类型准确表达具有较高的实际价值和应用前景。
[Abstract]:Because the cost of coring is high and the number of coring wells is small in Daqing Oilfield, it is very limited to judge the microfacies of sedimentary microfacies by analyzing coring samples. In this paper, using the logging curve data of well coring in oilfield, combining with expert knowledge, the pattern recognition method and application technology of neural network are studied, and the research results are applied to the sedimentary microfacies judgement of non-coring wells. To realize the evaluation of the oil reserves and exploitation value of a certain block. Combined with the actual analysis of sedimentary microfacies, it is found that, in terms of phase judgment data, besides the quantitative data related to logging curves, the qualitative information such as possible expert experience will also have an impact on the discriminant results. Moreover, the microphase type itself is fuzzy and random, so if only the quantitative recognition model is used to judge the phase, the error of the recognition result may be large. In order to solve the above problems, the fuzzy inference process neural network model and the mixed computing process neural network model based on cloud change are proposed to distinguish sedimentary microfacies, on the one hand, for the evaluation rule data with expert experience, the fuzzy inference process neural network model and the mixed computing process neural network model based on cloud variation are proposed. Fuzzy reasoning can be used to quantitatively process the qualitative information of logging facies in order to simplify the discriminant rules, extract effective discriminant data and improve the accuracy of sedimentary microfacies discrimination. On the other hand, for the data with fuzzy information, By using cloud model, the qualitative information in phase data can be transformed into the calculation process, which ensures the integrity and objectivity of the original data, considering the fact that the log phase data is a curve varying with depth. Therefore, the process input advantage of process neural network (PNN) is used to improve the accuracy of sedimentary microfacies discrimination by virtue of its ability to process the two dimensional information in time and space and to optimize the learning mechanism of process neural network continuously. Based on the data elements of sedimentary microfacies, the model and algorithm of sedimentary microfacies recognition are taken as the core. According to the development mode of software engineering, a sedimentary microfacies pattern recognition system is formed after software design and integrated development. Good experimental results are obtained. The research in this paper has a high practical value and application prospect for the actual processing of sedimentary microfacies discrimination data and the accurate expression of microfacies types.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TE319;TP183

【参考文献】