Integration of Monte-Carlo sampling method and deep-Q-learning network via COM-based interface between MATLAB/Python and Aspen Plus

Published: 14-10-2019| Version 1 | DOI: 10.17632/rz2zj86yzp.1
Soonho Hwangbo,


Latin hypercube sampling (LHS) method/Monte-Carlo (MC) simulation/Deep-Q-learning network (DQN) model Samples from the LHS method are generated in MATLAB and transferred to Aspen Plus to run MC simulation. Results from the LHS method and the MC simulation are utilized in the proposed DQN model. The agent in the DQN model is designed by Python and the environment in the DQN model is simulated in Aspen Plus. Actions from the agent in Python are transmitted to the environment in Aspen Plus. Results from the environment are sent back to the agent to improve the policy in the agent.