摘要
结合新型电缆地层测试器FCT的特点,将整个预测试过程简化为3个线性系统来反演地层渗透率。在算法中考虑了表皮效应和管线存储效应校正;编制了线性系统下的常规分析程序和非线性系统下的神经网络方法程序。并针对BP算法收敛速度慢的特点,引入Levenberg-Marquardt算法对BP模型进行改进,用实际压力数据进行测试,与Geoframe解释结果相比,吻合良好。通过实际验证,所建立的神经网络模型,预测精度良好,具有较好的非线性映射能力。
Combining the characteristic of Formation Characterization Tool （FCT）, a new type of wireline formation tester, the pressure changing pattern in whole pre-test process of FCT is simplified to three linear systems in formation permeability inversion model. Corrections for skin effect and pipeline storage effect are considered in the algorithm; a conventional analysis program with linear model and an artificial neural network （ANN） program with nonlinear model are developed; the Levenberg-Marquardt algorithm is introduced in the ANN model to improve the convergence rate of back propagation （BP） method. The algorithm and code test with real field pressure data results in a good agreement with the interpretation result of software Geoframe of Schlumberger. It shows that the ANN model has rather good forecast accuracy and nonlinear mapping ability.
出处
《测井技术》
CAS
CSCD
2007年第1期10-13,共4页
Well Logging Technology
基金
国家863项目(JSKF2004YJ14)资助
