Realtime Cost and Performance Improved Reservoir Simulator Service using ANN and Cloud Containers

Abstract

Real time reservoir simulation is growing demand while drilling to find new energy resources. Especially during drilling when the test data differs from actual data due to fault injections. This paper proposes a methodology using modified ANN scheduler using task characteristics and optimal cloud containers. Our methodology optimizes cost and end to end delay to achieve real time reservoir simulations. Realization of the paper is done using azure cloud resources and open porous media (OPM) reservoir simulator code. ANN based scheduling of cloud containers make the simulator energy efficient and scalable. Methodology uses microservice based architecture which gives the advantage of real time modifications, pluggability with minimum validation costs. Patent is demonstrated on 3-phase black oil well reservoirs - Input pod, Grid pod, Solver pods, Upscale pods, Output pods, 3D PODs. ANN scheduler with Ant Colony Optimization (ACO) will classify the input tasks based on task characteristics and schedule the POD containers on the optimal virtual machines (VMs). Proposed architecture is realized using Kubernetes docker containers on Microsoft azure linux VMs.