a hybrid cnn-lstm deep learning framework for porosity prediction in carbonate reservoirs

Porosity estimation in carbonate reservoirs is crucial for hydrocarbon exploration and production. traditional methods for porosity estimation, such as empirical correlations and geostatistical techniques, often rely on limited geological knowledge and may not capture the complex relationships between porosity and various geological features. this research aims to develop a novel hybrid convolutional neural network (cnn)-long short-term memory (lstm) deep learning framework for porosity prediction in carbonate reservoirs. the proposed framework aims to leverage the strengths of both cnns and lstms to capture both spatial and temporal patterns in well log data, thereby improving the accuracy and reliability of porosity estimation. the combination of cnns and lstms in a deep learning architecture allows for the efficient extraction of both local spatial features and long-term temporal dependencies in well log data. the model prediction performance improved from 0.67 (for mlp) to 0.98 for lstm, indicating the accuracy of the model. the results suggest that the cnn-lstm model can accurately estimate porosity in heterogeneous carbonate reservoirs, and its ability to capture spatial and temporal dependencies makes it well-suited for modeling complex geological systems.