Accurate Prediction Of Low-resistance Reservoirs Driven By Big Data

Since the 1950s,oil and gas resources have always occupied a dominant position in energy consumption.Therefore,it is of great significance to find the characteristics of reservoir and improve the recovery of reservoir.At present,with the deepening of oilfield exploration and development,conventional oil and gas reservoirs have been gradually exploited.Therefore,the exploration and identification of low resistivity oil and gas reservoirs has gradually become the focus of oil and gas exploration and development in mature oilfields.Due to the influence of many complex factors,the resistivity of low resistivity oil layer is obviously lower than that of conventional oil layer,and the logging response characteristics are not obvious,which reduces the ability of logging information to identify this kind of reservoir,and is often interpreted as water layer or even missed.Especially in the complex fault block reservoir,there are many factors such as sedimentary microfacies,structure,lithology,drilling and upper and lower layers.If we only rely on logging curve analysis to find oil layers,we will find that a large number of potential layers will be lost.At the same time,the interpretation conclusions of most sub-layers in the wellhead interpretation data are inaccurate,making it difficult to accurately predict lowresistance oil layers rich in a large amount of remaining oil.At present,more and more researchers choose to apply big data and data mining technology to efficiently mine valuable information from massive data to realize the potential of low resistivity reservoir.A number of experiments and studies have proved that the technology can effectively reduce the development cost and improve the recovery.Based on this background,this thesis proposes a multi-source data fusion method of sub-layer based on pattern recognition: select equal length logging curve data based on the change law of sub-layer logging curve,calculate cosine similarity of any two sub-layers for curve change pattern recognition,automatically set change pattern label for each sub-layer logging parameter combined with pattern recognition results,and fuse multiple logging curves based on sub-layer.The algorithm integrates logging,interpretation and production monitoring data into the sublayer data to realize the sub-layer data portrait.The example analysis shows that the algorithm reduces the calculation and storage cost of a large number of redundant logging data,and describes the multi-dimensional characteristics of each sub-layer more accurately and completely,which provides the data basis for the accurate identification of low resistivity reservoirs based on the fusion of multi-source data of sub-layer,the thesis analyzes the data characteristics of typical low resistivity oil layer,analyzes the oil-bearing.Association of typical low resistivity oil layer,calculates the similarity of typical oil layer,and deeply excavates the potential association relationship and parameter distribution interval between parameters of each sub-layer and oil-bearing property,so as to realize the target sub-layer screening.Based on the combination of the results and target layer selection,the intelligent recognition models of low resistivity oil layers such as Random forest,Adaptive boost tree and Gradient boosting tree are constructed to realize accurate potential tapping of low resistivity oil layers.The application of an oil field data in the eastern area is verified.The results show that the algorithm can accurately identify low resistivity oil layers,and the recognition accuracy is as high as 90%.Finally,the main work and innovation points of this thesis are summarized.In addition,the shortcomings of the research content and the future work are briefly put forward.