eduzhai > Applied Sciences > Engineering >

Deep Reinforcement Learning for Field Development Optimization

  • king
  • (0) Download
  • 20210507
  • Save

... pages left unread,continue reading

Document pages: 7 pages

Abstract: The field development optimization (FDO) problem represents a challengingmixed-integer nonlinear programming (MINLP) problem in which we seek to obtainthe number of wells, their type, location, and drilling sequence that maximizesan economic metric. Evolutionary optimization algorithms have been effectivelyapplied to solve the FDO problem, however, these methods provide only adeterministic (single) solution which are generally not robust towards smallchanges in the problem setup. In this work, the goal is to apply convolutionalneural network-based (CNN) deep reinforcement learning (DRL) algorithms to thefield development optimization problem in order to obtain a policy that mapsfrom different states or representation of the underlying geological model tooptimal decisions. The proximal policy optimization (PPO) algorithm isconsidered with two CNN architectures of varying number of layers andcomposition. Both networks obtained policies that provide satisfactory resultswhen compared to a hybrid particle swarm optimization - mesh adaptive directsearch (PSO-MADS) algorithm that has been shown to be effective at solving theFDO problem.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×