Recently, the team of Zhang Kai has made new progress in the field of reservoir automatic history matching, and the related research has been published in the SPE Journal. The paper is titled " Inversion Framework of Reservoir Parameters Based on Deep Autoregressive Surrogate and Continual Learning Strategy ".
Innovation: During the whole reservoir development cycle, continuous calibration of the reservoir parameters is necessary to adapt to new observations. Therefore, updating the surrogate model quickly to suit new data by transferring the experiences from previous tasks is essential. We establish a continual learning framework based on the ensemble strategy to address this problem. Leveraging this framework, we effectively reduce the computational cost of the surrogate model to learn new data. Although this framework can learn new data quickly, the accumulation of the error caused by the intrinsic limitation of the autoregressive model is unavoidable.
Abstract: 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.
The SPE Journal covers novel theories and emerging concepts (not including review articles or multi-part articles) spanning all aspects of engineering for oil and gas exploration and production, including drilling and completions, geomechanics, production and facilities, oilfield chemistry, CO2 sequestration and injection, reservoir evaluation and engineering, numerical simulation, data analytics, economics and externalities including health, safety, environment, and sustainability. The latest impact factor of the journal is 3.6, and the average impact factor IF in the past 3 years is 3.56. This journal currently appears in the JCR Q1 or the Chinese Academy of Sciences ranking 3.
Paper link:
https://doi.org/10.2118/215821-PA
Citation:
Zhang K., Fu W., Zhang J., et al. (2023) Inversion Framework of Reservoir Parameters Based on Deep Autoregressive Surrogate and Continual Learning Strategy[J]. SPE J. 28 (2023): 2223–2239.
