石油与天然气地质
石油與天然氣地質
석유여천연기지질
OIL & GAS GEOLOGY
2013年
3期
413-420
,共8页
艾宁%唐永%杨文龙%沈传波%王彦卿%黄文芳%尚婷
艾寧%唐永%楊文龍%瀋傳波%王彥卿%黃文芳%尚婷
애저%당영%양문룡%침전파%왕언경%황문방%상정
模糊化%神经网络%地震反演%登娄库组%长岭断陷
模糊化%神經網絡%地震反縯%登婁庫組%長嶺斷陷
모호화%신경망락%지진반연%등루고조%장령단함
fuzzy%neural network%seismic inversion%Denglouku Formation%Changling faulted depression
模糊神经网络储层反演,能够较好地摈弃不同类型、不同尺度数据的揉合,较好地保留了数据完整性,这既能反映整体的变化趋势,又能刻画局部细致变化特征。针对长岭1号气田登娄库组相变快、砂体横向连续性较差的特点,综合地震、测井、钻井测试分析资料,应用模糊神经网络原理,反演了研究区目的层段砂岩厚度与孔隙度特征。研究结果显示:①长岭1号气田登娄库组各段砂体平面展布变化较大,但厚度较大的区域均集中于长深103—长深1-3—长深1-1井和长深2井附近,向东北呈逐渐减薄趋势,D3岩层段长深103井周围岩石厚度可达33.2 m;②研究区登娄库组主力产油层D3和D4岩层段孔隙度较低,平均仅为5%左右,砂体孔隙度相对较高区域(>9%)仅集中于长深1井附近,整体显示登娄库组岩性较为致密;③反演结果与实际测试数据对比显示,砂体厚度误差控制在2.5 m以内,砂体孔隙度误差在0.49%以下,结果可靠性较好。利用模糊神经网络原理进行储层反演分析,能够很好地展现储集砂体分布规律及储集性能。
模糊神經網絡儲層反縯,能夠較好地擯棄不同類型、不同呎度數據的揉閤,較好地保留瞭數據完整性,這既能反映整體的變化趨勢,又能刻畫跼部細緻變化特徵。針對長嶺1號氣田登婁庫組相變快、砂體橫嚮連續性較差的特點,綜閤地震、測井、鑽井測試分析資料,應用模糊神經網絡原理,反縯瞭研究區目的層段砂巖厚度與孔隙度特徵。研究結果顯示:①長嶺1號氣田登婁庫組各段砂體平麵展佈變化較大,但厚度較大的區域均集中于長深103—長深1-3—長深1-1井和長深2井附近,嚮東北呈逐漸減薄趨勢,D3巖層段長深103井週圍巖石厚度可達33.2 m;②研究區登婁庫組主力產油層D3和D4巖層段孔隙度較低,平均僅為5%左右,砂體孔隙度相對較高區域(>9%)僅集中于長深1井附近,整體顯示登婁庫組巖性較為緻密;③反縯結果與實際測試數據對比顯示,砂體厚度誤差控製在2.5 m以內,砂體孔隙度誤差在0.49%以下,結果可靠性較好。利用模糊神經網絡原理進行儲層反縯分析,能夠很好地展現儲集砂體分佈規律及儲集性能。
모호신경망락저층반연,능구교호지빈기불동류형、불동척도수거적유합,교호지보류료수거완정성,저기능반영정체적변화추세,우능각화국부세치변화특정。침대장령1호기전등루고조상변쾌、사체횡향련속성교차적특점,종합지진、측정、찬정측시분석자료,응용모호신경망락원리,반연료연구구목적층단사암후도여공극도특정。연구결과현시:①장령1호기전등루고조각단사체평면전포변화교대,단후도교대적구역균집중우장심103—장심1-3—장심1-1정화장심2정부근,향동북정축점감박추세,D3암층단장심103정주위암석후도가체33.2 m;②연구구등루고조주력산유층D3화D4암층단공극도교저,평균부위5%좌우,사체공극도상대교고구역(>9%)부집중우장심1정부근,정체현시등루고조암성교위치밀;③반연결과여실제측시수거대비현시,사체후도오차공제재2.5 m이내,사체공극도오차재0.49%이하,결과가고성교호。이용모호신경망락원리진행저층반연분석,능구흔호지전현저집사체분포규률급저집성능。
Fuzzy neural network reservoir inversion can avoid the integration of data of different types and scales and well preserve the data integrity ,making it possible to reveal both the overall variation tendency and the local detailed features . In view of the rapid facies change and poor lateral continuity of the Denglouku Formation sandstone in the Changling 1 gas field,we integrated various data including seismic ,well logging,drilling and well testing and performed inversion of the sandstone thickness and porosity characters by using the fuzzy neural network theory .The following results were obtained .①The areal distributions of the Denglouku Formation sandstones in the Changling 1 gas field varies greatly ,with the thic-ker sandtone occurring mainly in areas along the Changshen 103,Changshen 1-3,Changshen 1-1 wells and the Changshen 2 well.The thickness of sandstone decreases progressively in northeast direction .The thickness of D3 layer near in the Changshen 103 well is up to 33.2m.②The major D3 and D4 pay zones in the Denglouku Formation have a relatively low porosity,averaging at only about 5%.The sandbodies with relatively high porosity (>9%)only occur in area around the Changshen 1 well.The Denglouku Formation is tight as a whole .③Comparison of the inversion results with the actual test data shows that the error of sandstone thickness is within 2.5 m and the error of the porosity is under 0.49%,indicating a high reliability of the inversion results .It is concluded that fuzzy neural network-based inversion of tight sandstone res-ervoirs can well reveal sandstone distribution and reservoir properties .