| Hydrocarbon-bearing reservoirs are a key focus of petroleum exploration and development.And accurately quantifying and characterizing such reservoirs through logging is a vital step towards understanding their characteristics,distribution,and scale.However,logging responses of hydrocarbon-bearing reservoirs become increasingly complex as the properties of the hydrocarbon in pores and their saturation state change.Furthermore,the petrophysical properties of reservoirs encountered during logging evaluations have become more complex,making it more challenging to achieve high accurate quantification of hydrocarbon-bearing reservoirs.This study systematically investigate the logging response characteristics and corresponding evaluation methods of shaly sand hydrocarbon-bearing reservoirs in the sequence E3d3 of LD oilfield.Firstly,the article presents a theoretical study of the conventional logging response characteristics in hydrocarbon-bearing reservoirs,including that of the neutron and density logging,resistivity logging,and spontaneous potential logging.The neutron and density logging combination exhibit a decrease in response values due to the decreasing of hydrocarbon density or flush zone water saturation.In addition,the attenuation of neutron logging response is further intensified by the excavation effect.In the case of resistivity logging,an increase in cation exchange capacity of hydrocarbon-bearing shaly sand leads to the decreasing of resistivity response values.For the spontaneous potential logging,its theoretical response model in shaly sand can be transformed into an integral equation of solution mass molar concentration.And the amplitude of spontaneous potential anomalies in shaly sand hydrocarbon-bearing reservoirs is significantly attenuated as the water saturation decreases.On the basis of clarifying the response characteristics of several conventional well logging methods,a series of quantitative evaluation models for shale volume,porosity,water saturation,and permeability are established for the hydrocarbon-bearing reservoirs in the sequence E3d3 of LD oilfield.The merit-selected models show good calculation results in both water and oil-bearing reservoirs.For gas-bearing reservoirs,a joint inversion method of porosity,water saturation,and hydrocarbon density based on particle swarm optimization is proposed and achieves good application results.In addition,the causes and mechanisms of low-resistivity oil and gas layers in the target interval are analyzed through petrophysics experiments and numerical simulations.The results show that the main causes include the influences of high bound water saturation,clay adsorption conductivity,and differences in formation water resistivity.Based on this,two new fluid property identification charts are proposed by taking into account the response characteristics of spontaneous potential logging in hydrocarbon-bearing shaly sand.The comprehensive interpretation accuracy of the new charts reach 93.5%,which are 16.67%higher than that of traditional density-resistivity charts.In addition,this study also systematically investigates the variation patterns of the nuclear magnetic resonance（NMR）logging response characteristics in hydrocarbon bearing reservoirs.The NMR transverse relaxation T2 distributions of eight water-wet samples with different pore structures under variation hydrocarbon saturation states are analyzed by numerical simulations.It is found that the T2 distributions and geometric means of the total signal,hydrocarbon signal,and water signal of NMR exhibit regular variations with changes in the water saturation of the samples.Based on these findings,a modified theoretical NMR model for characterizing the fluid signal response is proposed.And the reliability and superiority of the proposed model is verified by simulating and core NMR samples.Then the proposed model is successfully applied to the water saturation evaluation and NMR hydrocarbon correction in the sequence E3d3of LD oilfield.Additionally,from the data-driven perspective,a NMR logging hydrocarbon correction model is proposed based on the Seq2Seq recurrent neural networks.Comparative analysis of the model’s performance shows that the Seq2Seq-based correction method outperforms the existing two data-driven models. |
