Automating geobodies using insufficient labeled training data as input for structural prediction may result in missing important features and a possibility of overfitting, leading to low accuracy. We adopt a deep learning (DL) predictive modeling scheme to alleviate detection of channelized features based on classified seismic attributes (X) and different ground truth scenarios (y), to imitate actual human interpreters’ tasks. In this approach, diverse augmentation method was applied to increase the accuracy of the model after we were satisfied with the refined annotated ground truth dataset. We evaluated the effect of dropout as a training regularizer and facies’ spatial representation towards optimized prediction results, apart from conventional hyperparameter tuning. From our findings, increasing batch size helps speedup training speed and improve performance stability. Finally, we demonstrate that the designed Convolutional Neural Network (CNN) is capable of learning channelized variation from complex deepwater settings in a fluvial-dominated depositional environment while producing outstanding mean Intersection of Union (IoU) (95%) despite utilizing 6.4% from the overall dataset and avoiding overfitting possibilities.
SUMMARY
The key to further demonstrate 4D values in Field A relies on the interpretation of the reservoir changes due to production from year 1995 to 2006. Significant rock properties change due to pressure and fluid movement showed different amplitude variation with offset (AVO) behaviour on seismic data at different vintages. Existing AVO indicators suggest ambiguous results prone to certain factors such as porosity. Fluid properties such as oil and water of density has similar range of values. Hence, differentiating oil versus water in the porous media has become a problem typically in exploration prospecting and production monitoring. This problem can be mitigated if we are able to estimate accurately the bulk modulus parameters since it differs between the properties in the pore-fluid phase and rock-solid phase of homogeneous media. Since the conventional fluid factor (f) provides ambiguous results prone to the instability of pore fluid mixture and porosity in identifying fluid prospect, we re-parameterized new AVO approximation in term of K considering bulk modulus of the rock matrix higher than the bulk modulus of the effective pore fluid (combine with porosities higher than 15%).
Then, the inverse problem is solved using conjugate gradient optimization and tested the robustness of the new method on thin channelized synthetic model and real field data. The synthetic result shows the inverted pore fluid bulk modulus is closer to the measured model and the initial model, however larger uncertainties was observed from inverted density compared to the measured model. Starting from the noise free scenario, we gradually added noise from 5% to 30% into the channelized synthetic model to ensure the quality of the inversion method. Field data example revealed that 4D inverted effective pore-fluid bulk modulus are capable to identify lateral waterflood sweeping effect in the reservoir after eleven (11) years of production, expand the ability in differentiating fluid for clastic reservoir hence increase the probability for accurate hydrocarbon prediction.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.