The Scheduling of Dual Pipeline Crude Oil: A Reinforcement Learning Approach

In this paper, a short-term scheduling method for refinery crude oil in dual pipelines based on deep reinforcement learning algorithm is proposed. First of all, the short-term scheduling problem of dual-pipeline crude oil in refinery is converted into the solution of pipeline operation sequence. PPO algorithm is then used to construct action value network to solve the model. After that, we introduce the specific design of a refinery's dual pipelines based on deep reinforcement learning aiming at minimizing the total cost consumption. Finally, the action value network model is trained by using actual data, and the results are compared with the scheduling scheme of the currently used intelligent optimization algorithms. Simulation results indicate the feasibility and optimization of the proposed method, and provide a new approach for the short-term scheduling of dual pipeline crude oil in refineries.