Predicting Oil Movement in a Development System using Deep Latent Dynamics Models

Slides: https://bayesgroup.github.io/bmml_sem/2018/Temirchev_Metamodelling.pdf

We present a novel technique for assessing the dynamics of multiphase fluid flow in the oil reservoir. We demonstrate an efficient workflow for handling the 3D reservoir simulation data in a way which is orders of magnitude faster than the conventional routine. The workflow (we call it "Metamodel") is based on projecting the dynamical system into nonlinear subspace where the dynamics is captured by deep recurrent neural network. Compared to basic reduced order modelling approaches our projecting technique involves usage of variational autoencoder model instead of linear ones. We show that being trained on multiple results of the conventional reservoir modelling, the Metamodel does not compromise the accuracy of the reservoir dynamics reconstruction in a significant way. It allows forecasting not only the flow rates from the wells but also the dynamics of the distribution of pressure and fluid saturations within the reservoir. The results open a new perspective in the optimization of oilfield development as the scenario screening could be accelerated sufficiently.

During the talk, I will introduce you to the classical POD-Galerkin approach to reduce the computational cost of modelling multi-phase flows through a porous medium, to the recently published reduced order model based on POD and Deep Residual RNNs and to the Metamodelling technique proposed by us.
No background on oil field development routine is needed - I will make a small intro to the task.

Видео Predicting Oil Movement in a Development System using Deep Latent Dynamics Models канала BayesGroup.ru