SSA-BP神经网络在岩性识别中的应用研究

Research on application of SSA-BP neural network in lithology recognition

岩性识别是地层评价、油藏描述、测井解释等方面的一项重要工作内容。随着计算机计算能力的提升,学者们将人工智能算法应用到岩性识别领域中,并取得了不错的研究成果。首先,基于奇异谱分析对测井数据进行趋势分析,因为SSA对数据缺失具备很好地适合性,从而避免岩性识别出现错误判断;随后,对数据进行了主成分分析,将岩性识别的10种影响因素(不同测井数据)进行了降维处理,只保留了4种;最后,根据BP神经网络搭建了岩性识别模型,利用7000个4种不同岩性的测井数据进行训练,并使用整体数据剩余的60个岩性数据进行测试。测试结果显示,岩性识别结果和实际岩性相一致的占比约为93.3%,共有56个为正确识别,且多数神经网络输出的数值接近1,这说明该方法在岩性识别中具有较高的可靠性。

Lithology identification is an important work content in formation evaluation,reservoir description,logging interpretation and so on.With the improvement of computer computing power,scholars had applied artificial intelligence algorithms to the field of lithology identification,and achieved good research results.Firstly,trend analysis of logging data based on singular spectrum analysis was carried out,because SSA has good suitability for data missing,so as to avoid wrong judgment in lithology identification;10 influencing factors(different logging data)were dimensionally reduced,and only 4 were retained.Finally,a lithology identification model was built based on BP neural network,and 7000 logging data of 4 different lithologies were used for training,and used the remaining 60 lithological data from the overall data for testing.The test results showed that the proportion of lithology identification results that were consistent with the actual lithology was about 93.3%,a total of 56 were correctly identified,and most of the neural network output values were close to 1,this strongly indicates that the method had high reliability in lithology identification.