Traditional deep learning methods pursue complex and single network architectures without considering the petrophysical relationship between different elastic parameters.The mathematical and statistical significance o...Traditional deep learning methods pursue complex and single network architectures without considering the petrophysical relationship between different elastic parameters.The mathematical and statistical significance of the inversion results may lead to model overfitting,especially when there are a limited number of well logs in a working area.Multitask learning provides an eff ective approach to addressing this issue.Simultaneously,learning multiple related tasks can improve a model’s generalization ability to a certain extent,thereby enhancing the performance of related tasks with an equal amount of labeled data.In this study,we propose an end-to-end multitask deep learning model that integrates a fully convolutional network and bidirectional gated recurrent unit for intelligent prestack inversion of“seismic data to elastic parameters.”The use of a Bayesian homoscedastic uncertainty-based loss function enables adaptive learning of the weight coeffi cients for diff erent elastic parameter inversion tasks,thereby reducing uncertainty during the inversion process.The proposed method combines the local feature perception of convolutional neural networks with the long-term memory of bidirectional gated recurrent networks.It maintains the rock physics constraint relationships among diff erent elastic parameters during the inversion process,demonstrating a high level of prediction accuracy.Numerical simulations and processing results of real seismic data validate the eff ectiveness and practicality of the proposed method.展开更多
The three parameters of P-wave velocity, S-wave velocity, and density have remarkable differences in conventional prestack inversion accuracy, so study of the consistency inversion of the ＂three parameters＂ is very ...The three parameters of P-wave velocity, S-wave velocity, and density have remarkable differences in conventional prestack inversion accuracy, so study of the consistency inversion of the ＂three parameters＂ is very important. In this paper, we present a new inversion algorithm and approach based on the in-depth analysis of the causes in their accuracy differences. With this new method, the inversion accuracy of the three parameters is improved synchronously by reasonable approximations and mutual constraint among the parameters. Theoretical model calculations and actual data applications with this method indicate that the three elastic parameters all have high inversion accuracy and maintain consistency, which also coincides with the theoretical model and actual data. This method has good application prospects.展开更多
Seismic Rock physics plays a bridge role between the rock moduli and physical properties of the hydrocarbon reservoirs.Prestack seismic inversion is an important method for the quantitative characterization of elastic...Seismic Rock physics plays a bridge role between the rock moduli and physical properties of the hydrocarbon reservoirs.Prestack seismic inversion is an important method for the quantitative characterization of elasticity,physical properties,lithology and fluid properties of subsurface reservoirs.In this paper,a high order approximation of rock physics model for clastic rocks is established and one seismic AVO reflection equation characterized by the high order approximation(Jacobian and Hessian matrix)of rock moduli is derived.Besides,the contribution of porosity,shale content and fluid saturation to AVO reflectivity is analyzed.The feasibility of the proposed AVO equation is discussed in the direct estimation of rock physical properties.On the basis of this,one probabilistic AVO inversion based on differential evolution-Markov chain Monte Carlo stochastic model is proposed on the premise that the model parameters obey Gaussian mixture probability prior model.The stochastic model has both the global optimization characteristics of the differential evolution algorithm and the uncertainty analysis ability of Markov chain Monte Carlo model.Through the cross parallel of multiple Markov chains,multiple stochastic solutions of the model parameters can be obtained simultaneously,and the posterior probability density distribution of the model parameters can be simulated effectively.The posterior mean is treated as the optimal solution of the model to be inverted.Besides,the variance and confidence interval are utilized to evaluate the uncertainties of the estimated results,so as to realize the simultaneous estimation of reservoir elasticity,physical properties,discrete lithofacies and dry rock skeleton.The validity of the proposed approach is verified by theoretical tests and one real application case in eastern China.展开更多
基金supported by National Key R&D Program of China(2018YFA0702501)National Natural Science Foundation of China (41974140)+1 种基金Science and Technology Management Department,China National Petroleum Corporation(2022DQ0604-01)China National Petroleum Corporation-China University of Petroleum (Beijing) Strategy。
文摘Traditional deep learning methods pursue complex and single network architectures without considering the petrophysical relationship between different elastic parameters.The mathematical and statistical significance of the inversion results may lead to model overfitting,especially when there are a limited number of well logs in a working area.Multitask learning provides an eff ective approach to addressing this issue.Simultaneously,learning multiple related tasks can improve a model’s generalization ability to a certain extent,thereby enhancing the performance of related tasks with an equal amount of labeled data.In this study,we propose an end-to-end multitask deep learning model that integrates a fully convolutional network and bidirectional gated recurrent unit for intelligent prestack inversion of“seismic data to elastic parameters.”The use of a Bayesian homoscedastic uncertainty-based loss function enables adaptive learning of the weight coeffi cients for diff erent elastic parameter inversion tasks,thereby reducing uncertainty during the inversion process.The proposed method combines the local feature perception of convolutional neural networks with the long-term memory of bidirectional gated recurrent networks.It maintains the rock physics constraint relationships among diff erent elastic parameters during the inversion process,demonstrating a high level of prediction accuracy.Numerical simulations and processing results of real seismic data validate the eff ectiveness and practicality of the proposed method.
基金sponsored by the National Major Program (No. 2011ZX05006-006)the 973 Program of China (No. 2011CB201104)Technical Research of Elastic Flooding Boundary and Well Network Optimization at the Development Late Stage of Low Permeable Oil Field (No. 2011ZX05009)
文摘The three parameters of P-wave velocity, S-wave velocity, and density have remarkable differences in conventional prestack inversion accuracy, so study of the consistency inversion of the ＂three parameters＂ is very important. In this paper, we present a new inversion algorithm and approach based on the in-depth analysis of the causes in their accuracy differences. With this new method, the inversion accuracy of the three parameters is improved synchronously by reasonable approximations and mutual constraint among the parameters. Theoretical model calculations and actual data applications with this method indicate that the three elastic parameters all have high inversion accuracy and maintain consistency, which also coincides with the theoretical model and actual data. This method has good application prospects.
基金supported by the National Grand Project for Science and Technology(Grant Nos.2016ZX05024-004,2017ZX05009-001,2017ZX05036-005)the Science Foundation from SINOPEC Key Laboratory of Geophysics(Grant No.WTYJY-WX2019-0104)。
文摘Seismic Rock physics plays a bridge role between the rock moduli and physical properties of the hydrocarbon reservoirs.Prestack seismic inversion is an important method for the quantitative characterization of elasticity,physical properties,lithology and fluid properties of subsurface reservoirs.In this paper,a high order approximation of rock physics model for clastic rocks is established and one seismic AVO reflection equation characterized by the high order approximation(Jacobian and Hessian matrix)of rock moduli is derived.Besides,the contribution of porosity,shale content and fluid saturation to AVO reflectivity is analyzed.The feasibility of the proposed AVO equation is discussed in the direct estimation of rock physical properties.On the basis of this,one probabilistic AVO inversion based on differential evolution-Markov chain Monte Carlo stochastic model is proposed on the premise that the model parameters obey Gaussian mixture probability prior model.The stochastic model has both the global optimization characteristics of the differential evolution algorithm and the uncertainty analysis ability of Markov chain Monte Carlo model.Through the cross parallel of multiple Markov chains,multiple stochastic solutions of the model parameters can be obtained simultaneously,and the posterior probability density distribution of the model parameters can be simulated effectively.The posterior mean is treated as the optimal solution of the model to be inverted.Besides,the variance and confidence interval are utilized to evaluate the uncertainties of the estimated results,so as to realize the simultaneous estimation of reservoir elasticity,physical properties,discrete lithofacies and dry rock skeleton.The validity of the proposed approach is verified by theoretical tests and one real application case in eastern China.
