Abstract:Common-used method for gas reservoir detection by using seismic data is to pick up seismic attributes in fixed windows along the objective horizons and carry out attributes analysis,in which the oil/gas detection for thin-layer sandstone reservoir is more difficult.For that reason,the paper uses nonlinear random inversion method to finely describe the spatial distribution of sand body,getting the information of top and bottom horizons in reservoir;then picks up various seismic attributes along the horizons and windows change with the thickness variation of reservoir;and finally carries out gas-bearing formation detection by using RS theory-based attribute optimized method and SOM neural network mode recognition method respectively.The method was used in Luodai gasfield of Chuanxi depression,resulting in 85% and above of matched rate.