Аннотация:A semi-automated 3D genetic inversion has been used for reservoir property prediction in the Shtokman gas/condensate field. The basic input requirements for the workflow are a post-stack seismic cube and relevant logs at some control wells. A non-linear neural network operator is designed to transform the seismic cube into an acoustic impedance cube. Intelligent data reduction and a true 3D approach allows to derive a stable solution for the given inversion problem, whilst limiting the negative effects of under-and over-estimation in the prediction procedure. Crossplots, curve fitting and Gaussian simulation techniques are used to populate the property model with reservoir parameters. In the Shtokman field three Jurassic pay zones have been defined. Other reservoir parameters like density, hydrocarbon saturation and gamma-ray have also been genetically inverted. The results are compared with the conventional geological model, as accepted by the Russian State Committee for Reserves. The latter is based on interpreted seismic horizons and data obtained from conventional attribute analysis. The genetic inversion earth model for the Shtokman field has a higher resolution and hence is considered of better quality than the existing field reservoir model; it can be used in the further field development planning efforts.