【摘要】 将遗传算法、Fisher最优分割和边缘检测算法相结合,设计了一种适用于多参数测井地层划分的遗传最优分割算法.该算法以测井测点号为基因,以有序测点号组为染色体,由测点的边缘隶属度和段间/段内离差总和的比值共同构成综合优化指标(适应度),采用锦标赛选择机制,通过MonteCarlo抽样进行交叉(通过两两染色体比较设定抽样区间),采用灾变变异和精英保持策略,利用多参数测井数据求取最优化地层分界线.在济阳坳陷中生界地层划分中的实际应用,表明该方法的自动分层结果符合地质实际,且计算速度很快.该方法对于没有露头、岩心稀少、地震勘探分辨率低和地球物理测井数据量大情况下的深部隐伏地层划分有很好的适用性.更多还原

【Abstract】 In this paper we propose a genetic algorithm for optimal stratigraphical division using multi-parameter well-log data.The algorithm incorporates the Fisher optimal division and edge detection methods into genetic searching in order to find out rapidly from a well a group of division boundaries that satisfy an optimization criterion.The optimization criterion involves the following steps:After a well(or a segment of it)is divided into a group of subsections,the total in-subsection deviation of the multi-parameter log data is minimized.Then the total inter-subsection deviation of the data is maximized,and the boundary points are to the largest extent singular points in log curves.This criterion is used to formulate the fitness function in our genetic algorithm.In the algorithm,logging measurement points are taken as genes,and groups of ordered measurement points as chromosomes;Genetic strategies of tournament selection,catastrophic mutation,Monte Carlo sampling crossover,and elitist reservation are adopted.We test the method on a geologically well-studied well which penetrates the deep-buried Mesozoic strata in Jiyang depression,Shandong,China.It is shown that,using log data alone,our method obtains stratigrahic division results that are geologically sound.Meanwhile the method proves far faster than an ordinary Fisher optimal division method in computation.We conclude that our method is highly applicable in statratigraphic division studies in areas such as Jiyang depression,where the interested strata is deep-buried,with scarce relevant well cores,poor seismic resolution,and abundant log data.更多还原