Tag: salt tectonics

A Procedure to Parameterize High Permeability Zones in Naturally Fractured Reservoir

Publicado em por .

This paper presents a novel-methodology to compensate for the poor characterization of high-permeability structures (excess-K: vugs, karsts and super-K features), and non-fault-related-fractures, in naturally fractured Brazilian Pre-Salt carbonate reservoirs. These heterogeneities are often undetectable in well logs and seismic data, but significantly impact well performance. The methodology aims to enhance the representation of such features within dynamic simulation models, improving reservoir characterization and supporting more reliable data-assimilation and forecasting processes. The methodology involves: (1) upscaling high-fidelity fine-grid models to coarser-grids while preserving dynamic behavior, (2) identifying wells with productivity/injectivity mismatches due to a poor excess-K characterization, (3) applying a data assimilation (DA) process to minimize the mismatch between modeled and measured wells production and injection rates by updating the absolute permeability of the matrix. The novelty of the process is that the permeability field is updated by creating a mask (3D property) built by kriging permeability increments estimated from the well cells with productivity/injectivity issues. Therefore, the DA aims to find the least increments of permeability needed for each well such that when this mask is summed with the matrix permeability field all wells present good productivity/injectivity matching with history data. The methodology was applied to a dual-porosity/dual-permeability (DP/DK) compositional reservoir model. Two distinct well behaviors were observed: (1) wells located within fracture zone (12 of 33) showed good productivity/injectivity alignment with historical data and (2) the remaining 21 wells, located away from fracture zone, exhibited significantly poorer productivity/injectivity. This mismatch was attributed to the absence of excess-K features in the original matrix permeability model (Km-field). The optimization process was applied to these 21 wells. For each well a specific Ki-value was settled, defining input-points for kriging. The resulting kriged permeability correction volume (mask) was summed with the Km-field to generate an updated-permeability model. This process was repeated until all wells presented good productivity/injectivity matching with historical data. The process not only corrected the simulated dynamic responses, but also revealed key spatial permeability patterns that had not been captured in the static model. The results served as feedback to the geologists and enabled iterative improvement of the geological model, supporting a more integrated and realistic characterization. Overall, the results validate the methodology as a robust tool for incorporating unresolved high-permeability features in reservoir simulation and improving the quality of data assimilation. This study introduces an automated, iterative probabilistic data-assimilation framework that directly integrates geostatistical kriging with permeability adjustments for excess-Kstructures. The approach provides bidirectional feedback to geological modeling and allows the generation of realistic ensembles for data assimilation workflows. By combining geo-statistics within an uncertainty reduction scheme, the method addresses key modeling gaps encountered when modelling a Brazilian Pre-Salt carbonate.

Estimating Rock Typing in Uncored Wells Using Machine Learning Techniques for Brazilian Pre-Salt Carbonate Reservoir

Publicado em por .
Accurate rock typing in uncored wells is essential for enhancing reservoir models, particularly in complex geological formations like the Brazilian pre-salt carbonate reservoirs. This study explores the application of machine learning (ML) techniques to estimate rock types in uncored wells. The research leveraged core data from 11 cored wells to calculate the Rock Quality Index (RQI) and Flow Zone Indicator (FZI), identifying 14 distinct rock types through the Discrete Rock Typing (DRT) method, along with well log data such as gamma ray, density, neutron, permeability, and porosity. Various machine learning models, including Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Decision Tree, Gradient Boosting, Naive Bayes, XGBoost, and Random Forest were tested, where XGBoost achieved the highest accuracy of 73.3%. Applying XGBoost to all wells resulted in accuracy ranging from 0.6 to 0.91 and the model was subsequently used to estimate rock types in over 20 uncored wells to generate reservoir simulation models. This study highlights the efficacy of machine learning (particularly XGBoost) in addressing reservoir complexities and offering significant improvements in the understanding and development of carbonate reservoirs.