Wensheng Zhou

Wensheng Zhou

Auto-generated profile, based on publicly available information

Wensheng Zhou

CNOOC Research Institute Ltd

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
2 June 2023 in SPE Journal
Verified
Journal Article
Closed access
PDF file
Kai ZhangKai Zhang
Kai Zhang
WFWenhao Fu
Wenhao Fu
JZJinding Zhang
Jinding Zhang
WZWensheng Zhou
Wensheng Zhou
CLChen Liu
Chen Liu
JYJun Yao
Jun Yao
Summary: History matching is a crucial process that enables the calibration of uncertain parameters of the numerical model to obtain an acceptable match between simulated and observed historical data. However, the implementation of the history-matching algorithm is usually based on iteration, which is a computationally expensive process due to the numerous runs of the simulation. To address this challenge, we propose a surrogate model for simulation based on an autoregressive model combined with a convolutional gated recurrent unit (ConvGRU). The proposed ConvGRU-based autoregressive neural network (ConvGRU-AR-Net) can accurately predict state maps (such as saturation maps) based on spatial and vector data (such as permeability and relative permeability, respectively) in an end-to-end fashion. Furthermore, history matching must be performed multiple times throughout the production cycle of the reservoir to fit the most recent production observations, making continual learning crucial. To enable the surrogate model to quickly learn recent data by transferring experience from previous tasks, an ensemble-based continual learning strategy is used. Together with the proposed neural network–based surrogate model, the randomized maximum likelihood (RML) is used to calibrate uncertain parameters. The proposed method is evaluated using 2D and 3D reservoir models. For both cases, the surrogate inversion framework successfully achieves a reasonable posterior distribution of reservoir parameters and provides a reliable assessment of the reservoir’s behaviors.

21 March 2024 in SPE Journal
Verified
Journal Article
Closed access
PDF file
CCChenming Cao
Chenming Cao
XXXiaoming Xue
Xiaoming Xue
Kai ZhangKai Zhang
Kai Zhang
LSLinqi Song
Linqi Song
Liming ZhangLiming Zhang
Liming Zhang
CLChen Liu
Chen Liu
Summary: Production optimization is a crucial component of closed-loop reservoir management, which typically aims to search for the best development scheme for maximum economic benefit. Over the decades, a large body of algorithms have been proposed to address production optimization problems, among which the surrogate-assisted evolutionary algorithm (SAEA) gained much research popularity due to its problem information-agnostic implementation and strong global search capability. However, existing production optimization methods often optimize individual tasks from scratch in an isolated manner, ignoring the available optimization experience hidden in previously optimized tasks. The incapability of transferring knowledge from possibly related tasks makes these algorithms always require a considerable number of simulation runs to obtain high-quality development schemes, which could be computationally prohibitive. To address this issue, this paper proposes a novel competitive knowledge transfer (CKT) method to leverage the knowledge from previously solved tasks toward enhanced production optimization performance. The proposed method consists of two parts: (1) similarity measurement that uses both reservoir features and optimization data for identifying the most promising previously solved task and (2) CKT that launches a competition between the development schemes of different tasks to decide whether to trigger the knowledge transfer. The efficacy of the proposed method is validated on a number of synthetic benchmark functions as well as two production optimization tasks. The experimental results demonstrate that the proposed method can significantly improve production optimization performance and achieve better optimization results when certain helpful previously optimized tasks are available.