Lithologically Controlled Pore Pressure Prediction and Geomechanical Modelling Using Probabilistic Multivariate Clustering Analysis and an Expert System

Curtis, Alan A. (eGAMLS Inc.) | Eslinger, Eric (eGAMLS Inc.) | Nickerson, Randy (Caza Petroleum) | Nookala, Siva (Cerone Pvt Ltd) | Boyle, Francis (eGAMLS Inc.)

OnePetro 

Abstract

A workflow is presented which places far greater emphasis on formation lithology than is usually employed during pore pressure and geomechanical studies. Advanced classification techniques are linked with conventional pore pressure prediction and geomechanical modelling methods to implement the new workflow. The lithological classifications which are developed permit more robust predictions by facilitating the constraint of pore pressure and geomechanical results to available well data. Lithological assignments are developed from well logs using a Bayesian-based multivariate clustering analysis technique which yields a probabilistic Electroclass at each depth along the wellbore. The probabilistic results are analysed with an Expert System that automatically assigns a Lithology to the Electroclass at each depth. The Expert System can be modified for different regions and adjusted (and overruled) by an experienced analyst. The resulting multivariate model, with probabilistic lithological assignment, is used to QC, and if necessary predict, well log curves in missing intervals along the wellbore. Thus, interval velocities across the complete well profile from surface to total depth can be established from well log sonic data. These lithology-dependent velocities are then used to develop pore pressure predictions using an effective stress method in which the governing parameters are themselves lithologically dependent. Likewise, geomechanical properties such as Poisson's Ratio, Young's Modulus, Brittleness Index, and the minimum horizontal stress are calculated using Lithology-dependent parameters. An example is presented for an onshore US unconventional formation in which multiple wells are used to develop a robust lithological classification. The developed lithology then controls the wireline log curve predictions and ultimately the pore pressure and geomechanical predictions in selected wells. The impact of different lithologies on pore pressure and geomechanical estimates can be clearly seen and the impact of parameter setting ascertained for each. It is concluded that predictions of pore pressure and geomechanical properties are considerably enhanced by the far better understanding and consistent inclusion of lithology.