Data-driven production optimization using particle swarm algorithm based on the ensemble-learning proxy model

Shu-Yi Du,Xiang-Guo Zhao,Chi-Yu Xie,Jing-Wei Zhu,Jiu-Long Wang,Jiao-Sheng Yang,Hong-Qing Song
DOI: https://doi.org/10.1016/j.petsci.2023.04.001
IF: 4.757
2023-04-08
Petroleum Science
Abstract:Production optimization is of significance for carbonate reservoirs, directly affecting the sustainability and profitability of reservoir development. Traditional physics-based numerical simulations suffer from insufficient calculation accuracy and excessive time consumption when performing production optimization. We establish an ensemble proxy-model-assisted optimization framework combining the Bayesian random forest (BRF) with the particle swarm optimization algorithm (PSO). The BRF method is implemented to construct a proxy model of the injection–production system that can accurately predict the dynamic parameters of producers based on injection data and production measures. With the help of proxy model, PSO is applied to search the optimal injection pattern integrating Pareto front analysis. After experimental testing, the proxy model not only boasts higher prediction accuracy compared to deep learning, but it also requires 8 times less time for training. In addition, the injection mode adjusted by the PSO algorithm can effectively reduce the gas–oil ratio and increase the oil production by more than 10% for carbonate reservoirs. The proposed proxy-model-assisted optimization protocol brings new perspectives on the multi-objective optimization problems in the petroleum industry, which can provide more options for the project decision-makers to balance the oil production and the gas–oil ratio considering physical and operational constraints.
energy & fuels,engineering, petroleum
What problem does this paper attempt to address?