测井数字处理中最优化方法的选择

本文提出了在测井数字处理中评价与选择最优化数学方法的准则,根据多矿物解释模型建立了最优化测井解释数学模型,采用四组实际测井值(每组又用两种储层参数初始值),对常用的九种最优化方法进行了系统的考察分析。根据144种不同组合的计算结果的对比分析说明,用抛物线插值法作一维搜索与BFGS变尺度法作多维搜索的最优化方法处理测井资料,收敛速度快,占用机时少,对初始点要求较宽,数值稳定性好,优化解释结果好,是比较适合于测井数字处理与储集评价的最优化方法,这已由10多口井的最优化测井数字处理结果所证实。

SELECTION OF THE OPTIMIZATION METHOD FOR DIGITAL WELL LOGGING DATA PROCESSING

This paper studies the selection of apprapriate optimization methods for digital well logging data processing and oil reservoir evaluation. The criteria for evaluation and selection of optimization mathematical methods in digital logging data processing were suggested. Based on the mathematical model of optimization log interpretation from the multimineral model, the commonly used nine optimimization methods were systematically examined with four groups of logging data and two sets of initial points values and comparison of 144 groups of data show that the BFGS optimization method with parabolic linear search is the most suitable for logging data processing. This method has the advantages of faster convergence, short computation time and good numerical stability. Logging data processing from more than 10 wells proved that this method requires less initial point selection and gives the better results of optimization log interpretation.