A New Porosity Prediction Method Based on Deep Learning of TabNet Algorithm

Porosity is one of the core parameters in the process of understanding subsurface fluid flow, reservoir characterization, and reservoir evaluation. Due to the limitation of the number of cores taken, conventional experimental analysis can only obtain a small amount of porosity data. How to improve the accuracy of porosity prediction has always been an issue. It is one of the hotspots in the study of reservoir parameter prediction. This paper proposes a porosity prediction method based on Tab Net, and compares it with traditional machine learning and LSTM porosity prediction methods. The results show that the RMSE of the three porosity prediction models are 3.48,3.67, and 3.95 respectively. Therefore, this paper believes that the model based on the TabNet network can more effectively predict the reservoir porosity and provide a new method for reservoir evaluation ideas and methods.