论文部分内容阅读
长庆气田马五１ 储层非均质严重，储层种类多且各类储层的分布规律及特征信息模糊不清，缺乏先验知识与经验模式的特点。本项研究引进了模糊神经网络方法，利用测井数据建立了长庆气田马五１ 储层测井判识的模糊神经网络。计算结果表明，其辨识精度（ 回判） 达到了９７ ．４５ ％ ，为利用测井进行储层划分提供了一种方便迅速的工具。经过气田近７０ 口气井的储层分类验证，取得了较为令人满意的结果，表明此分析技术在特定背景条件下较之单一的模糊模式识别或 Ｂ Ｐ 神经网络的精度更高，方法更先进，它既可以避免测井储层识别中因环境影响岩性、电性特征模糊而引起的误差，还可以避免因噪音数据影响网络学习限入局部极小的可能，且网络在接受了特征数据后学习速度也得到较大提高。
The Ma5-1 reservoir in Changqing gas field is characterized by serious heterogeneity, many reservoir types, and the distribution and characteristic information of various types of reservoirs are vague and lack the characteristics of prior knowledge and experience mode. In this study, a fuzzy neural network method was introduced, and a well logging data was used to establish a fuzzy neural network for identification of Ma5-1 reservoir in Changqing gas field. The calculation results show that the recognition accuracy (return) reaches 97. 45%, providing a quick and easy tool for well logging using well logging. Through the verification of reservoir classification in nearly 70 gas wells in the gas field, satisfactory results have been obtained, indicating that this analytical technique is more accurate and has more advanced methods than the single fuzzy pattern recognition or BP neural network under certain background conditions , Which not only avoids the error caused by environmental lithology and electrical characteristics fuzzy in logging reservoir identification, but also avoids the possibility that the network learning will be limited to a local minimum due to the noise data, and the network has accepted the characteristic data After learning speed has also been greatly improved.