摘要
火山岩化学成分变化大、矿物组成复杂,利用测井资料识别火山岩岩性较困难。为了提高松辽盆地火山岩岩性识别的准确率,为火山岩气藏测井精细评价奠定基础,以火山岩岩石学特征为指导,总结了松辽盆地火山岩测井响应规律,得到了对岩性敏感的测井参数:密度、自然伽马、铀、钍、钾等;并利用这些参数绘制了交会图,建立了岩性识别图版;引入支持向量机方法,建立了火山岩岩性识别模型,并利用网格搜索技术优化了模型参数,最优模型的精度和推广能力均优于交会图版。将支持向量机岩性识别模型应用于松辽盆地实际井资料的处理中,将处理结果与岩石薄片鉴定结果对比,符合率高,证明了该模型的准确性。
Volcanic rock was featured with complex mineral composition and changeable chemical compositions,therefore,it was difficult for lithologic identification of the volcanic rock with logging data.In order to improve the accuracy of its lithologic identification,and lay a foundation for fine logging evaluation of volcanic gas reservoirs in Songliao Basin,based on petrologic characteristic study of volcanic rocks,the influential laws of well logging in volcanic rock were summed up,the logging parameters sensitive to lithology were obtained,such as density,natural gamma,uranium(U),thorium(TH)and potassium(K)and et al.These parameters were used to develop crossplots for establishing a chard board for lithologic identification.A method of support vector machine(SVM)was introduced to establish a model for volcanic lithologic identification,the parameters of the model was optimized by using technique of network searching.The accuracy and generalization of the optimal model were better than those of the crossplots.The SVM-based model is used to process logging data of wells in Songliao Basin,and its computation results are well consistent with the core thin section,it is proven that the model is accurate.
出处
《石油天然气学报》
CAS
CSCD
2014年第3期72-76,6-7,共5页
Journal of Oil and Gas Technology
关键词
火山岩
岩性识别
测井参数
交会图
支持向量机
松辽盆地
lithologic identification
logging parameter
crossplot method
support vector machine(SVM)
Songliao Basin