Integrating Geostatistics and Machine Learning Techniques for 3D Electrofacies Modeling in Early Cretaceous Clastic Reservoirs, Lower Indus Basin, Pakistan

This study focuses on characterising the heterogeneous Lower Goru Formation (Early Cretaceous) in Pakistan's Lower Indus Basin using advanced machine learning techniques. Due to limited core data, traditional petrophysical methods struggle to accurately identify facies. The research uses electrofacies analysis by integrating well logs and clustering algorithms—specifically K-Means and Self-Organizing Maps (SOM)—to classify facies and determine reservoir properties.

After validating the approach on a well with available drilled facies, the SOM proved more accurate than K-Means, identifying seven distinct facies. 3D facies models, integrated with seismic and petrophysical data, revealed better reservoir quality in the eastern and central parts of the study area.

The study introduces a robust workflow for reservoir characterization, improving the accuracy of facies identification and aiding in better well planning by overcoming limitations of conventional methods through machine learning and geostatistical integration.