Abstract: In recent years,thin reservoir is one of the main targets of oil and gas exploration in China. To enhance the ability of thin reservoir identification can be significant for detecting the residual hydrocarbon resources. Due to the limitation of seismic resolution and the interference of seismic reflection among neighboring thin layers,it is a tricky task to directly evaluate thin reservoir based on the seismic reflection data. Therefore,a new method of seismic reflectivity inversion based on sparse Bayesian learning was adopted,and then the relative acoustic impedance of underground strata was estimated by the inverted seismic reflectivity. A linear FIR filtering was also introduced to eliminate the low-frequency accumulative errors which were generated in the processing of relative impedance calculation. Thin reservoir can be finally identified based on the filtered relative impedance with higher confidence. Field application shows that the accuracy of seismic reflectivity can be improved significantly by seismic reflectivity inversion based on sparse Bayesian learning,laying a solid foundation for a better estimation of relative impedance of underground formation. The designed linear FIR filter can effectively suppress the low-frequency accumulative errors in the estimated relative impedance,improving the accuracy of thin reservoir detection. Comparing with the traditional seismic amplitude,the relative impedance estimated by the proposed method can reveal the lateral distribution of thin reservoir more accurately,and finally improves the success rate of thin reservoir exploration.

Key words: thin reservoir, sparse Bayesian learning, seismic reflectivity, FIR high-pass filtering, relative impedance