Refinement of a 3D geological model through neural-simulation-based seismic prediction

Prediction of sand reservoir properties plays a key role in the exploration and development of oil and gas fields. Traditional approaches often face limitations associated with nonlinear functions, heterogeneities, and variability of rocks. These challenges lead to a decrease in the accuracy of net reservoir prediction, which entails risks in reservoir engineering and field development. In this regard, the implementation of machine learning methods that can automatically identify complex patterns, take into account multi-factor relationships, and adjust to changing conditions becomes relevant, which opens up new opportunities to improve the predicting accuracy and reliability. This paper discusses modern neural prediction methods, their advantages and disadvantages, as well as practical aspects of applying machine learning to predict sand reservoirs. Particular attention is paid to the selection of input data, creation of neural network architecture, setting up estimation parameters, and interpreting the results. The study is aimed at demonstrating the high performance of neural network technologies in solving problems of predicting the sand reservoir properties. It is expected that the results of the study will contribute to the optimization of geological exploration and improve the economics of field development.