Enhancing Gas Reservoir Characterization by Simulated Annealing Method (SAM)

Abstract

Recently Ouenes etused extensively simulated annealing to generate permeability and porosity fields. In this paper, the Simulated Annealing Method (SAM) is applied to a gas storage reservoir. The permeability field predicted in four geological layers preserves the spatial distribution and honors the actual measured data from the field. These known data at the wells lead to isotropic and anisotropic variograms which were matched by SAM. In this approach, contrasting the usual conditional simulation, no interpolation technique is used to smooth data at unknown locations. The generated field is compared to a kriged field by using a 3D advanced gas reservoir simulator. When using permeability fields generated by SAM, gas and water pressure data obtained through simulation become very close to observed field pressure data. On the other hand, the pressure obtained with the kriged field clearly shows that the smooth representation introduced by the kriging is not realistic, and the heterogeneities which affect considerably the flow behavior are not well described.

Physics‐Informed Convolutional Decoder (PICD): A Novel Approach for Direct Inversion of Heterogeneous Subsurface Flow

“…Non‐gradient‐based methods offer the advantage of circumventing the direct computation of gradient information, making them more adaptable to integration with numerical simulators. Within this category, intelligent optimization algorithms stand out as heuristic methods, which have been broadly used in solving inverse problems, such as Genetic Algorithm (GA) (Sayyafzadeh & Haghighi, 2012; Xavier et al., 2013), Simulated Annealing (SA) (Jeong & Park, 2019; Ouenes et al., 1992), and Particle Swarm Optimization (PSO) (S. Lee & Stephen, 2019; Mohamed et al., 2010). Markov chain Monte Carlo (MCMC) methods constitute another subset of non‐gradient‐based methods, offering the posterior distribution of the model parameters through sampling from the desired probability distribution.…”

Reservoir characterisation using artificial bee colony optimisation

“…Various optimisation methods have been used in history matching, such as simulated annealing (Ouenes et al, 1993), genetic algorithm (GA) (Romero and Carter, 2001), particle swarm (Mohamed et al, 2010), ant colony (Hajizadeh et al, 2011) and several classical optimisation methods-for example: Gauss-Newton, Levenberg-Marquardt and Limited-memory, Broyden-Fletcher-Goldfarb-Shanno (LBFGS) (He et al, 1997, Zhang et al, 2005. Each method has its own capabilities and weaknesses, and the method selection should be based on conditions-especially the shape of landscape.…”