Field Development Planning (FDP) for oil and gas (O&G) recovery projects requires strategic well placement and infrastructure design to maximize hydrocarbon extraction. This work presents a Deep Reinforcement Learning (DRL) framework based on Deep Q-Networks (DQN) for the optimization of drilling scheduling. Different sets of parameters describing the reservoir characteristics are considered for the definition of the state space and the parameter combination providing the largest profit over the entire lifetime of the oil project is identified. The proposed framework is validated considering a synthetic case study that emulates the complexity of the drilling scheduling problem in real-world scenarios.
Deep Reinforcement Learning for Strategic Asset Management in Oil & Gas Recovery Projects
Abdin A.;Baraldi P.;Pinciroli L.;Zio E.
2024-01-01
Abstract
Field Development Planning (FDP) for oil and gas (O&G) recovery projects requires strategic well placement and infrastructure design to maximize hydrocarbon extraction. This work presents a Deep Reinforcement Learning (DRL) framework based on Deep Q-Networks (DQN) for the optimization of drilling scheduling. Different sets of parameters describing the reservoir characteristics are considered for the definition of the state space and the parameter combination providing the largest profit over the entire lifetime of the oil project is identified. The proposed framework is validated considering a synthetic case study that emulates the complexity of the drilling scheduling problem in real-world scenarios.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
