基于KNN-Fisher算法的测井解释知识库构建方法

包含专家经验知识的知识库对复杂测井解释评价有着重要作用,但测井解释知识库普遍存在自动调用准确度不高的问题。因此将KNN算法(K邻近法)与Fisher判别法相结合,提出了一种基于KNN-Fisher算法的测井解释知识库构建方法。该方法利用地区经验模型及专家经验参数等研究成果建立知识库,通过混合推理查找知识,对唯一确定的知识直接调用。如果存在多条知识,则利用KNN-Fisher判别法调用知识。方法具体流程如下:首先通过Fisher判别法建立知识模型,然后将需要调用知识的层组测井资料导入模型,再通过引入权重系数并利用质心点约束的K邻近法对知识进行判别调用,最终实现知识的自动调用。将该方法应用于南海西部海域油气田测井解释评价,结果表明,利用该方法自动获取到的知识与所在地层条件吻合,该方法可有效克服混合推理和KNN算法存在的不足,实现测井解释知识库的准确调用。

Conventional knowledge bases for log interpretation offer low accuracy in automated calls.To address this issue,a method to construct the knowledge base relying on KNN-Fisher,a combination of an improved k-nearest neighbors (KNN) algorithm and Fishers discriminant,is proposed.In this approach,a regional empirical model and expert-judged empirical parameters constitute the raw input.The method seeks knowledge through hybrid reasoning to produce unique,direct calls.In case of uncertain knowledge,the following procedure is adopted:first,the knowledge model is established through Fishers discriminant; then,the logging data of the layer group for which knowledge is sought is imported into the model; finally,the information is discriminated using the KNN algorithm constrained by a weight coefficient and a centroid point.This results in an automatic knowledge call.The method has been applied to the log interpretation of a region in the western part of the South China Sea.The results show that the knowledge acquired by the algorithm was consistent with the actual stratum conditions.It is concluded that the proposed method can effectively overcome the shortcomings of hybrid reasoning and KNN as it produces accurate calls from a knowledge base for log interpretation.

中海石油综合科研项目(YXKY-2019-ZJ-06)资助。