Permeability Predictions for Tight Sandstone Reservoir Using Explainable Machine Learning and Particle Swarm Optimization

High-precision permeability prediction is of great significance to tight sandstone reservoirs. However, while considerable progress has recently been made in the machine learning based prediction of reservoir permeability, the generalization of this approach is limited by weak interpretability. Hence, an interpretable XGBoost model is proposed herein based on particle swarm optimization to predict the permeability of tight sandstone reservoirs with higher accuracy and robust interpretability. The porosity and permeability of 202 core plugs and 6 logging curves (namely, the gamma-ray (GR) curve, the acoustic curve (AC), the spontaneous potential (SP) curve, the caliper (CAL) curve, the deep lateral resistivity (RILD) curve, and eight lateral resistivity (RFOC) curve) are extracted along with three derived variables (i.e., the shale content, the AC slope, and the GR slope) as data sets. Based on the data preprocessing, global and local interpretations are performed according to the Shapley additive explanations (SHAP) analysis, and the redundant features in the data set are screened to identify the porosity, AC, CAL, and GR slope as the four most important features. The particle swarm optimization algorithm is then used to optimize the hyperparameters of the XGBoost model. The prediction results of the PSO-XGBoost model indicate a superior performance compared with that of the benchmark XGBoost model. In addition, the reliable application of the interpretable PSO-XGBoost model in the prediction of tight sandstone reservoir permeability is examined by comparing the results with those of two traditional mathematical regression models, five machine learning models, and three deep learning models. Thus, the interpretable PSO-XGBoost model is shown to have more advantages in permeability prediction along with the lowest root mean square error, thereby confirming the effectiveness and practicability of this method.

Managing Sanding Risk in Sandstone Reservoir Through a New Constitutive Model

Sandstone reservoir failure during hydrocarbon production can cause negative impact on the oil/gas field development economics. Loss of integrity and hydrocarbon leakage due to downhole or surface erosion can decrease the risk of operational safety. Therefore, a proper understanding of the best formulation to manage and find the balance between productivity and sand risk is very important. Making decisions for the best and most economical completion design needs a full and proper sanding risk analysis driven by geomechanics modeling. The accuracy of modeling the reservoir rock mechanical behavior and the failure analysis depends on the selection of the constitutive model (failure criteria) specially to understand the failure and post failure mechanisms. Thus, an appropriate constitutive model/criterion is required as most of the current model/criteria are not developed for a weak rock material honoring the non-linearity and post failure (softening) process. Therefore, a new and novel elasto-plastic constitutive model for sandstone rock has been investigated and developed. The effort started with a sequence of triaxial tests at different confining pressures on core samples. Different types of rock have been tested during the developing and validation of the constitutive model. Comparison with other existing failure criteria was also performed. As the results, the newly developed constitutive model is better honoring the full spectrum of elasto-plastic rock mechanical behavior (softening and post-failure) which is important for oil and gas applications, specifically for sand production and drilling i.e. failure stabilization due to stress relief. The formulation and process are demonstrated with a case study for an old gas field, where a few gas wells have been shut-in due to severe sand production. The sand production predictive models have been validated with downhole pressure. The wells have been side-tracked and recompleted using the new sand failure prediction, using the new formulation resulted in restoring sand-free production at former rates. The novelty of this study would be in finding the right formula to best design the predictive model and to avoid any sand production when using the newly developed constitutive model.