Identification And Application Of Steam Breakthrough Channels After Multiple Cycles Of Huff And Puff In Heavy Oil Reservoirs

After several rounds of steam huff and puff production in Well 3711 of Jinglou Oilfield in Henan,problems such as a large number of steam channeling channels and a sharp decrease in oil-steam ratio occurred.In order to improve the development effect,it is urgent to identify and analyze the steam breakthrough channels,and switch the appropriate reservoir development method to control the steam channeling and improve the recovery rate of the oil reservoir.In this thesis,the development dynamics and steam channeling laws are firstly analyzed,and the causes of steam channeling are studied.Then,the mechanism model of steam channeling in the target work area is established.Based on the above factors,a machine learning feature attribute set is constructed.After data reconstruction,similarity analysis,dimensionality reduction,and random oversampling are performed on the feature set,11 feature attributes are finally obtained,and a training sample set and a steam breakthrough channel reconstruction set are established.Then,the random forest regression algorithm is used to construct the identification model of steam channeling,which can reconstruct the relative steam channeling degree field and predict the future development of steam channeling in oil wells.In order to verify the accuracy of the machine learning model and explore the feasibility of the steam channeling control method,this thesis established a numerical model of the actual oilfield,and verified that the temperature field and oil saturation field of the numerical model and the relative steam channeling degree field reconstructed by machine learning have good coupling The prediction accuracy of the machine learning model for oil well gas channeling is high.Based on the reconstructed relative steam channeling degree field and numerical simulation steam channeling field,this thesis studies the adaptation of different steam channeling control methods,such as grouping huff and puff,steam flooding with deformed well pattern,steam channel blockage and transfer to conventional steam flooding,etc.The results show that: based on the steam channeling degree field reconstructed by machine learning,the nitrogen foam-assisted steam flooding in the form of reverse nine-point mining has good effect in controlling steam channeling,and the final recovery rate is 53.95%.Based on reservoir engineering,machine learning,reservoir numerical simulation and other disciplines,this thesis analyzes the influencing factors of steam channeling,establishes an intelligent identification and analysis model of steam channeling channels,studies the adaptability of different steam channeling control measures,and helps improve steam injection in target reservoirs.It plays an important role in the development effect and guiding the practice of mineral production.