All-in-one proxy to replace 4D seismic forward modeling with machine learning algorithms

Danaei, Shahram; Cirne, Marcos Vinícius Mussel; Maleki, Masoud; Schiozer, Denis José; Rocha, Anderson; Davolio, Anderson

doi:10.1016/j.geoen.2023.211460

Cited by 5 publications

(7 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The ensemble of models consists of 100 uncertain dynamic models. As also highlighted by [18], it is noteworthy that the ensemble of models used here were not submitted to any data assimilation process. To prepare the data used to train the proxy model, three average maps of each property (porosity, NTG, and saturation/pressure changes) were generated to more accurately appraise the dimension of the reservoir.…”

Section: Unisim-iv Descriptionmentioning

confidence: 99%

“…Lastly, [18] proposed a new methodology for map-based all-in-one proxy (S4D-Proxy) to fully replace the traditional seismic forward modeling using ML algorithms (Extreme Gradient Booster -XGBoost and Deep Neural Network -DNN). The S4D-Proxy model consists of a ML map-based proxy that can predict the root-mean-square amplitude (dRMS) difference using reservoir properties as input features (porosity, net-to-gross ratio -NTG, and saturation and pressure variations).…”

Section: Introductionmentioning

confidence: 99%

“…Therefore, alternatives using proxy models were proposed to circumvent this problem and optimize this process. [18] reviewed some of the main applications of proxy models which are mostly used in the inverse problems to estimate saturation-pressure changes from 4D seismic data. However, the application of proxy models in 4D seismic history matching to replace the traditional 4D seismic forward modeling is limited to some studies.…”

Section: Introductionmentioning

confidence: 99%

“…The S4D-Proxy provided results with high fidelity in comparison to the ground truth (synthetic seismic data generated from the traditional forward modeling), which were evaluated using the coefficient of determination R 2 -score. In [18], the optimal outcomes were achieved by employing the DNN architecture. This architecture will therefore serve as the reference baseline for our proposed approaches.…”

Section: Introductionmentioning

confidence: 99%

“…In [18], the authors emphasized that although the S4D-Proxy can provide successful results in a sandstone reservoir case, in more complex contexts such as carbonate reservoirs, e.g., UNISIM-IV benchmark [20], the proxy may not provide good predictions and might require changes related to the ML algorithm structure and training set increase. Carbonate reservoir complexity can be appreciated in the PEM formulation for these rocks, as observed in [21].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Efficient proxy for Time-Lapse Seismic ForwardModeling using U-Net Encoder-Decoder Approach

Diniz,

Maleki,

Cirne

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

The time-lapse seismic (4D seismic) forward modeling provides crucial data for calibrating reservoir models through the reservoir data assimilation method. Unfortunately, conventional 4D seismic forward modeling methodology is time-expensive and entails significant computational resource consumption. To address these drawbacks, in this work, our goal is to develop a proxy model for the 4D seismic forward modeling using a class of machine learning algorithm named U-Net encoder-decoder. We applied the developed proxy model to a benchmark carbonate reservoir using an ensemble of reservoir simulation models from UNISIM IV (a synthetic benchmark based on real data of a Brazilian pre-salt field). Moreover, we aim to introduce seminal strategies for interpreting the proposed proxy model operation, its outputs, and possible correlations between input and output variables. To achieve this, we trained and tested two versions of U-net-based models and applied methods for explainable AI, such as Grad-CAM and Forward Feature Selection. The experiments showed good results when applied to the test dataset. The R-squared values were in the range of 0.7 to 0.9, showing the efficiency of the proxy model to replace the 4D seismic forward modeling. Additionally, the qualitative analysis helped us identify which input properties and regions of the reservoir are more relevant for the model's inference. These results are a valuable preliminary step toward a robust, explainable machine learning-based proxy forward modeling.

show abstract

Section: Unisim-iv Descriptionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%