Advanced Reservoir Permeability Distribution Prediction: A Discrete Cosine Transform-Based History Matching Workflow

Abstract Reliable reservoir simulation is essential for effective reservoir management in the petroleum industry, as it enables engineers to make well-informed decisions on production strategies and resource allocation. However, accurately calibrating reservoir models to match observed field data remains a significant challenge. This difficulty primarily arises from the inherent uncertainties in key parameters such as permeability and porosity, which are crucial for predicting reservoir behavior and performance. Traditional history matching methods, which involve adjusting reservoir parameters to align simulated models with observed production data, often require substantial computational resources and can produce multiple solutions that do not necessarily reflect the true reservoir conditions. These methods typically rely on extensive trial and error, leading to models lacking geological accuracy and realism. To overcome these limitations, there is increasing interest in applying advanced mathematical techniques and intelligent algorithms that offer more efficient, precise, and geologically consistent solutions. This paper introduces a novel workflow for enhancing the accuracy and efficiency of history matching in reservoir simulations using real-field data. The approach addresses uncertainties in permeability and porosity relationships, particularly when modifying relative permeability (RP) might conflict with lab-measured data. A multi-objective Differential Evolution algorithm is employed to tackle these uncertainties, reduce randomness, and identify optimal solutions along the Pareto front. This process is followed by Principal Component Analysis (PCA), used for clustering all optimal solutions along the Pareto front, leading to several clusters of well-matched oil/water (O/W) models. Finally, the Discrete Cosine Transform (DCT) is applied to capture key reservoir features and expedite optimization. By converting the permeability array into the frequency domain, the DCT achieves significant data compression while preserving essential geological characteristics, thereby reducing data complexity and enhancing the efficiency and accuracy of history matching. The methodology culminates in integrating the matched O/W models with bottom-hole pressure (BHP) data, resulting in comprehensive O/W/BHP matched models. This framework minimizes uncertainties and provides an efficient, systematic approach for improving history matching in complex reservoir simulations.