Abstract:The thickness prediction of thin layer or thin-interbedded strata is an important part of lateral reservoir prediction. Conventional methods for thin-layer thickness estimation are achived by deriving single parameter from time-domain or frequency-domain seismic data. This method consists of following essential steps:·Do maximum entropy analysis in time-frequency domain by making wavelet transform.·Extract multiple characteristic parameters.·Compute thickness of thin layer or thin-interbedded strata by using nonlinear relations among seismic characteristic parameters which are sensitive to thinlayer thickness,and by taking neural network algorithm.The forward and inversion results of theoretic model show that this method,being of some noise resistant ability,brings good effects in predicting both thin-layer thickness and accumulative thickness of thin-interbedded strata. This method was used to predict the accumulative thickness of thin-interbedded sand strata in Carboniferous reservoir of partial DL92-04 seismic section in TZ area.The prediction is satisfactory.