The application of artificial intelligence seismic inversion techniques in sandstone and shale reservoir studies

Artificial Intelligence Probabilistic Neural Networks are a type of neural network based on the theory of probability density functions, widely applicable in areas such as pattern recognition. Addressing the current issue of low prediction accuracy in seismic lithology inversion, the authors propose a specific probabilistic neural network method, including the construction of network models and steps for prediction and identification. The primary target layer in the study area is the lower part of the Sha-1 Member of the Shaximiao Formation, characterized by lakeshore and river channel sands. The reservoirs are thin-layered with significant lateral lithological variations, making conventional reservoir prediction using seismic data challenging. Therefore, leveraging the correlation between well log responses, seismic attributes, and geological lithological characteristics of the reservoirs in the area, the probabilistic neural network method is employed to transform seismic attribute data for predicting and identifying stratigraphic features. This approach has achieved good results in practical data processing.