CPC E21B 44/00 (2013.01) [E21B 47/026 (2013.01); E21B 49/003 (2013.01); E21B 49/0875 (2020.05); E21B 2200/20 (2020.05); E21B 2200/22 (2020.05)] | 21 Claims |
1. A method for drilling a well, comprising:
obtaining an observable from a working environment, wherein the working environment comprises an algorithmic component in a reinforcement learning framework, wherein the working environment further comprises a representation of current drilling operations of a well;
generating, by a plurality of working agents in the working environment, a plurality of proposed drilling actions for the well in response to the observable;
synchronizing a plurality of validation agents in a validation environment with the plurality of working agents in the working environment, wherein the validation environment initially represents the working environment;
simulating, for the well, drilling responses to the proposed drilling actions using a plurality of validation agents in the validation environment, wherein the simulating comprises performing multiple simulations, for each of a plurality of priorities, for each of the plurality of proposed drilling actions, wherein the plurality of priorities comprise: drilling efficiency, drilling speed, risk of failure, and planned trajectory adherence;
determining rewards for each of the plurality of proposed drilling actions based on the simulating, using the validation agents;
selecting one of the plurality of proposed drilling actions based on the rewards;
providing the selected one of the plurality of actions to the working environment; and
causing a drilling rig to execute the selected one of the proposed actions.
|