Well Placement Optimization in a Mature Carbonate Waterflood using Streamline-based Quality Maps

Abstract

The placement of infill producers and injectors is an important aspect in the overall development strategy of any field and is particularly challenging for mature fields with high levels of water-cut. Previous screening approaches based upon static reservoir quality maps have limited applicability as these do not account for the drainage and swept volumes from existing wells. In contrast, direct application of formal optimization methods such as evolutionary algorithms and adjoint-based methods to high resolution geologic models may better represent reservoir dynamics but can be complex to implement or computationally prohibitive. We propose a novel method for well placement optimization that relies on streamlines which represents the flow paths in the reservoir and the time of flight which represents the travel time of fluids along streamlines. Specifically, the streamline time of flight from the injectors provides swept volumes for injectors whereas streamline time of flight from producers gives drainage volumes for producers. These quantities can be effectively combined to a ‘total time of flight’ to locate the potential regions of unswept and undrained oil in the reservoir. Our approach utilizes a dynamic measure based on the total streamline time of flight combined with static parameters to identify potential locations for infill drilling. Areas having high value of the dynamic measure (sweet spots) are both poorly drained and poorly swept, making them attractive for drilling infill wells. We show the power and utility of our proposed method on a mature offshore carbonate field in western India. The simulation model was history matched using a hierarchical history matching approach that follows a sequence of calibrations from global to local parameters in coarsened and fine scales. Using our proposed method on the history matched model we obtained a dynamic measure map highlighting areas suitable for drilling infill wells. Finally, we compared the performance of infill wells located using the dynamic measure map with wells located using traditional well placement techniques, for example, oil saturation map from simulation. Our proposed method consistently outperforms the traditional approaches. Subsequent field infill drilling in the field has validated our approach.

Two-Step Predict and Correct Non-Intrusive Parametric Model Order Reduction for Changing Well Locations Using a Machine Learning Framework

“…Thus, as we saw above for POD, the main aim is to replace the computationally expensive solver in Equation (6) with a reduced form of the Jacobian and Residual, as shown in Equation (24), where r N d . To this end, we aim to construct reduced-order models or reduced-order basis that are an accurate approximation of the fine-scale solution in the entire parameter domain-i.e., for any ζ in Equation (3).…”

“…Thus, training a PMOR model must be tied to the optimization strategy used and training samples must be designed accordingly. For example, since well location optimization is a very challenging problem given numerous possible well configurations, researchers resort to quick evaluations or diagnostics of the reservoir to rule out many areas in the reservoir as potential well locations [24,40]. Then, given this information, complex optimization routines using simulation runs can be sought for a systematic decision-making process.…”

“…Well location optimization in the context of surrogate or low-complexity models was developed in References [23,24]. However, until now, for projection-based reducedorder models the focus mainly has been on developing them for well-control optimization problems.…”

Well placement optimization for offshore oilfield based on Theil index and differential evolution algorithm

“…Well placement optimization problem is usually complex, nonconvex and multimodal (Awotunde and Naranjo 2014;Wang et al 2016). In order to solve well placement optimization problem, both gradient-based and derivativefree optimization algorithms are applied (Yeten et al 2002;Bangerth et al 2006;Taware et al 2012;Bouzarkouna et al 2012;Nwankwor et al 2013;Awotunde 2014;Isebor et al 2014;Humphries and Haynes 2015;Carosio et al 2015;Khademi and Karimaghaee 2016;Bagheri and Masihi 2016), such as finite difference gradient (FDC), simultaneous perturbation stochastic approximation (SPSA), genetic algorithm (GA), particle swarm optimization (PSO) algorithm, covariance matrix adaptation evolution strategy (CMA-ES) algorithm and differential evolution (DE) algorithm. Due to the simple metaheuristic mechanism, DE algorithm has shown good performance in solving the real parameter optimization problem and global optimization problem (Carosio et al 2015).…”

Optimization of Well Placement and Production for Large - scale Mature Oil Fields

“…These fields have considerable potential because they have generally been developed for many years under suboptimal conditions. Development adjustments, such as drilling new infill wells and changing the injection/production rates of wells, can improve the performance of mature fields and significantly increase oil recovery [2]. Infill wells are new wells added to an existing field within original well patterns.…”