一基于Bayesian多分支岩石可钻性值估计-中国石油大学.PDF

2014年摇 第38卷摇 摇 摇 摇 摇 摇 中国石油大学学报(自然科学版)摇 摇 摇 摇 摇 摇 摇 摇 Vol.38摇 No.3 摇 第3期摇 摇 摇 摇 摇 摇 摇 摇 Journal of China University of Petroleum摇 摇 摇 摇 摇 摇 摇 摇 Jun.2014 摇 摇 文章编号:1673鄄5005(2014)03鄄0073鄄07摇 摇 摇 doi:10.3969/ j.issn.1673鄄5005.2014.03.012 一基于 Bayesian多分支岩石可钻性值估计 1 2 1 1 沙林秀 ,邵小华 ,张奇志 ,李摇 琳 (1.西安石油大学陕西省钻机控制重点实验室,陕西西安710065;2.大庆钻井集团钻井一公司,黑龙江大庆710072) 摘要:针对智能优化控制过程中岩石可钻性参数估计存在非实时性和模型泛化能力差的问题,采用两层结构建立基 于Bayesian多分支岩石可钻性估计模型。 通过Bayesian分类器实现岩性分类以提高可钻性模型样本数据的相关性, 细化可钻性估计模型;采用改进双链量子遗传算法优化的BPNN结构,根据不同的岩石类型建立相应的岩石可钻性 IDCQGA_BPNN估计模型。 结果表明,该方法通过算法优化网络模型增强了模型的泛化能力,加快了参数的估计速 度和估计精度,能够满足智能优化控制过程中岩石可钻性参数估计的实时性需求。 关键词:岩石可钻性;Bayesian分类器;L-M算法;改进的双链量子遗传算法 中图分类号:TP 183摇 摇 摇 文献标志码:A Estimation of rock drillability based on a Bayesian multi鄄branch model 1 2 1 1 SHA Lin鄄xiu ,SHAO Xiao鄄hua ,ZHANG Qi鄄zhi ,LI Lin (1.Key Laboratory of Drilling Rigs Controlling Technique,Xian Shiyou University,Xian710065,China;謖 謖 2.The First Drilling Company,Daqing Drilling Corporation,Daqing 710072,China) Abstract:A two鄄level model was establishedfor predictingrocksdrillability basedon a Bayesian multi鄄branch model in or鄄謖 der to improve the real鄄time calculating capability of the model and increase its generalization ability for intelligent optimiza鄄 tion control. By using the Bayesian method for lithology classification,the correlations of different rock samples and their drillability can be refined,and consequently the rock drillability model can be improved. Using an optimized back鄄propaga鄄 tion neural network(BPNN)with animproveddouble鄄chainquantumgeneticalgorithm(IDCQGA),thenew model of IDCQ鄄 GA_BPNN can be modified according to the lithology typ