Deep Reinforcement Learning for Simulation Model Optimization

The proposed work leverages advanced deep reinforcement learning techniques, specifically Deep Q-Learning (DQN), by combining Q-learning with neural networks to optimize VRFB-specific parameters, ensuring a strong fit between real and simulated data. This method outperforms existing approaches in voltage prediction. Furthermore, we enhance the approach by incorporating a second deep RL algorithm : Dueling DQN an improvement over DQN, which yields a 10% enhancement in results, particularly in voltage prediction. The final approach produces an accurate VRFB model that can be generalized to various types of redox flow batteries.