基于堆栈自编码模型的岩性分类方法

Lithologic classification method based on stack self-coding model

基于测井资料进行岩性分类是油气勘探和采集过程中的基础问题之一。根据前人的研究成果已开发了多种岩性分类方法,提出了基于堆栈自编码模型的岩性分类方法,通过对测井资料进行逐层抽象提取特征凸显不同岩性的微弱测井响应,使用Softmax作为分类器进行岩性分类。应用该方法对丁山地区泥质灰岩和灰质泥岩进行识别,将训练好的模型应用于丁山HF1井中,正确率均达到94.39%以上。

The lithology classification based on the log information is one of the basic problems in the process of oil and gas exploration and acquisition.The lithology classification method based on stack self-coding model have been developed with the previous research results.The log information are abstracted in layer-by-layer for the characteristics to highlight the weak log responses of different lithology.The Softmax is used as a classifier to do the lithology classification.The method proposed in this paper is used to identify the argillaceous limestone and limestone mudstone in Dingshan area.The trained model is applied to Well HF1 of Dingshan,the result show that the correct rate is above 94.39%.