作　　者： (李民）; (陈科贵）; (杨智）; (张家浩）; (刘鑫）;

机构地区： 西南石油大学地球科学与技术学院,四川成都266100

出　　处： 《测井技术》 2017年第4期453-457,共5页

摘　　要： 利用常规测井曲线、交会图等方法识别稠油油藏岩性正确率低,效率不高。选用已采集的油田测井资料,分别采用费歇判别法、BP神经网络、极限学习机3种模式识别法,对稠油油藏岩性进行识别,并利用岩心、录井资料对识别结果进行检验。研究结果表明,费歇识别法正确率为81.2%,BP神经网络识别正确率为90.3%,极限学习机识别正确率为92%。在合理有效提取参数的条件下,极限学习机和BP神经网络的岩性识别符合率要高于费歇判别法,且具有运算速度快、效率高等优点。模式识别法在测井解释岩性识别方面有一定的推广应用价值。 Lithology identification is important in log interpretation.Identifying lithology accurately with logging data is beneficial to improve the development efficiency.Since there is low accuracy and efficiency for the conventional logging curves and cross plot to identify lithology of heavy oil reservoir, we use Fisher Different Approach, BP Neural Network and ELM to identify lithology with the suitable logging data.The results show that the accuracy of Fisher Different Approach is 81.2%, BP Neural Network is 90.3%, ELM is 92%.Under the condition of reasonable parameters, ELM and BP Neural Network have higher accuracy than Fisher Different Approach, and they also have fast operation and high efficiency.Pattern recognition has some application value in lithology identification.

关 键 词： 测井解释 模式识别 岩性识别 稠油油藏