METHOD AND DEVICE FOR PREDICTION OF RESERVOIR PARAMETERS BASED ON LIMITATION OF GEOLOGICAL CHARACTERISTICS, AND DATA STORAGE MEDIUM Russian patent published in 2025 - IPC G01V20/00 G01V1/28 G01V1/30 G06N3/44 G06N3/45 G06N3/88 G06F30/27 

FIELD: geophysics.

SUBSTANCE: invention discloses a geological property-based method for predicting reservoir parameters, a device, a computer storage medium and a computer device. Method includes: selection of dominant seismic attributes in accordance with relevance between different types of seismic attributes of target formation and reservoir parameters (S100); based on the dominant seismic attributes, classifying the seismic waveforms of the target formation using a predetermined waveform classification network model and in accordance with the waveform features to obtain a waveform classification result (S200); acceptance of waveform classification result as constraint, and construction of different models of deep neural network corresponding to different geological characteristics (S300); training of various deep neural network models based on seismic and target reservoir logging data in order to optimize the model parameters of each of the deep neural network (S400) models; combining different trained deep neural network models into a set of spatially varying neural network prediction models (S500); and prediction of reservoir parameters of target formation using a set of spatially varying neural network prediction models (S600).

EFFECT: high accuracy and information content of the obtained data.

13 cl, 13 dwg