中国石油大学学报(自然科学版)
中國石油大學學報(自然科學版)
중국석유대학학보(자연과학판)
JOURNAL OF CHINA UNIVERSITY OF PETROLEUM(EDITION OF NATURAL SCIENCE)
2015年
3期
188-194
,共7页
戴永寿%岳炜杰%孙伟峰%李立刚%张亚南%程佳成
戴永壽%嶽煒傑%孫偉峰%李立剛%張亞南%程佳成
대영수%악위걸%손위봉%리립강%장아남%정가성
井喷%溢流监测%专家系统%改进的贝叶斯判别%“三高”油气井
井噴%溢流鑑測%專傢繫統%改進的貝葉斯判彆%“三高”油氣井
정분%일류감측%전가계통%개진적패협사판별%“삼고”유기정
blowout%kick detection%expert system%improved Bayes discriminant%"three high" wells
为解决钻井过程中传统溢流监测实时性和可靠性差的问题,设计开发出一套“三高”油气井早期溢流在线监测与预警系统。该系统针对常规单一参数、单一手段溢流监测实时性低、可靠性差的问题,综合应用微流量、随钻压力( pressure while drilling, PWD)以及综合录井3类参数,通过多参数、多手段相互印证的方式提高溢流监测的实时性与可靠性。提出基于专家系统和改进的贝叶斯判别相融合的溢流等钻井事故判别方法:当缺少训练数据时,应用专家系统判别溢流等钻井事故;当训练数据充足时,应用专家系统和改进的贝叶斯判别相结合识别溢流等钻井事故。改进的贝叶斯判别模型对于因属性变量间不独立而引起的误判具有一定的抑制作用,能够提高判别准确度,且训练简单,应用灵活。通过两种算法的有机结合,优势互补,可以提高井控安全的智能化水平与现场适用性。测试实验结果表明,该系统能够实时、有效地监测溢流等钻井事故。
為解決鑽井過程中傳統溢流鑑測實時性和可靠性差的問題,設計開髮齣一套“三高”油氣井早期溢流在線鑑測與預警繫統。該繫統針對常規單一參數、單一手段溢流鑑測實時性低、可靠性差的問題,綜閤應用微流量、隨鑽壓力( pressure while drilling, PWD)以及綜閤錄井3類參數,通過多參數、多手段相互印證的方式提高溢流鑑測的實時性與可靠性。提齣基于專傢繫統和改進的貝葉斯判彆相融閤的溢流等鑽井事故判彆方法:噹缺少訓練數據時,應用專傢繫統判彆溢流等鑽井事故;噹訓練數據充足時,應用專傢繫統和改進的貝葉斯判彆相結閤識彆溢流等鑽井事故。改進的貝葉斯判彆模型對于因屬性變量間不獨立而引起的誤判具有一定的抑製作用,能夠提高判彆準確度,且訓練簡單,應用靈活。通過兩種算法的有機結閤,優勢互補,可以提高井控安全的智能化水平與現場適用性。測試實驗結果錶明,該繫統能夠實時、有效地鑑測溢流等鑽井事故。
위해결찬정과정중전통일류감측실시성화가고성차적문제,설계개발출일투“삼고”유기정조기일류재선감측여예경계통。해계통침대상규단일삼수、단일수단일류감측실시성저、가고성차적문제,종합응용미류량、수찬압력( pressure while drilling, PWD)이급종합록정3류삼수,통과다삼수、다수단상호인증적방식제고일류감측적실시성여가고성。제출기우전가계통화개진적패협사판별상융합적일류등찬정사고판별방법:당결소훈련수거시,응용전가계통판별일류등찬정사고;당훈련수거충족시,응용전가계통화개진적패협사판별상결합식별일류등찬정사고。개진적패협사판별모형대우인속성변량간불독립이인기적오판구유일정적억제작용,능구제고판별준학도,차훈련간단,응용령활。통과량충산법적유궤결합,우세호보,가이제고정공안전적지능화수평여현장괄용성。측시실험결과표명,해계통능구실시、유효지감측일류등찬정사고。
In order to solve the problem about the poor instantaneity and reliability of early kick detection during drilling process by traditional methods, a realtime monitoring and warning system for kick foreboding on "three high" wells was de-signed and developed. Regarding the above defects of kick detection by conventional means in which only one parameter or one mode was used, the method makes use of three parameters which include micro-flux, pressure while drilling ( PWD) and comprehensive logging parameters, and applies the multi-parameter and multi-means to cross-verify and improve the instanta-neity and reliability of kick detection. An identification method based on expert system in combination with the improved Bayes discriminant was proposed. If lacking of training data, the method applies expert system to identify kick and other ac-cidents;otherwise it applies expert system combining with the improved Bayes discriminant to identify kick and other acci-dents. The improved Bayes discriminant model can reduce the misjudgment caused by the attribute variables which are not independent completely, and thus improve the identification accuracy. The application is more flexible, and the training is easier. The combination of the two kinds of intelligent algorithms, whose advantages are complementary can improve signifi-cantly the intelligent level and applicability. The verification tests show that the system can detect kick and other accidents instantaneously and effectively.