New paper entitled "Learning to Calibrate Battery Models in Real-Time with Deep Reinforcement Learning"
In our recent paper, we propose a Reinforcement Learning (RL) based calibration for the battery models. Our RL based approach does not need the ground truth degradation parameters for training, but learns by tracking the internal state of the battery.
Our experiments show that the RL generalizes better than directly training neural networks on the ground truth couplets of state and degradation parameters. Also, the RL based approach works faster than the Unscented Kalman Filter (UKF) based approach while providing more stable tracking of the degradation parameters.
The Chair of Intelligent Maintenance Systems focuses on developing intelligent algorithms to improve performance, reliability and availability of complex industrial assets and making the maintenance more cost efficient. For more information visit our webpage