The quality of an oil and gas field development project depends greatly on the accuracy of forecasting the processes that occur in the pore space of reservoirs during the extraction of hydrocarbons under certain technological conditions in production wells. The forecasting is possible if there is a geological model of the field. The more detailed the model is, the more accurate the prediction will be. The whole amount of information used to create a geological model of a field is of discrete nature, and its level of detail is determined by the number of wells that have discovered pay formations. One of the most important elements of the geological model is the nature of changes in reservoir properties of productive formations along their stretch and perpendicular to bedding. The creation of elements of this type requires information from laboratory studies of core material, interpretation of the wells logging results and methods for predicting the nature of changes in reservoir properties in the interwell space. The presence of these elements makes it possible to investigate the situation in which sedimentation (within the existing wells) took place and what types of facies the geological sections of the drilled producing intervals correspond to. Lithofacial zoning of the productive formation according to this information makes it possible to trace the regularities of distribution of facies of various types, to establish their mutual location, and accordingly to predict the nature of changes in reservoir properties in the interwell space. The lack of a sufficient amount of core material is a typical problem that makes it difficult to identify facies. There is another way to solve this problem – this is the identification of facies according to the morphology of logging curves. Nowadays, this problem is solved at a qualitative level. In this paper, it is proposed to apply a quantitative method for identifying facies using an artificial neural network. In particular, the morphology of curves is formalized by a number of parameters that form the input vector of an artificial neural network. At the output of the network, the clusters of logging curves with a similar morphology are formed. The authors refer these clusters to a certain type of facies analytically. On the basis of the information obtained, lithofacial zoning of the productive formations is carried out