Data-Driven Induction Motor Control by Reinforcement Learning Including a Model Predictive Safety Layer / by Felix Book ; Supervisor: Dr.-Ing. Oliver Wallscheid, Barnabas Haucke-Korber, Maximilian Schenke. Paderborn, 2022
Inhalt
- Introduction
- Motivation and goal of the thesis
- Results of the system model identification
- Structure of the thesis
- Analysis of the different algorithms in the hyperparameter optimization
- Drive System
- Reinforcement Learning Agent
- Mathematical formulation of the problem
- RL algorithms
- Deep Q Networks algorithm
- Rainbow algorithm
- Proximal Policy Optimization algorithm
- Evolution Strategies
- Implementation of the RL agent
- Flux Estimator
- Model Predictive Safety Layer
- Hyperparameter Optimization
- Implementation of the hyperparameter optimization
- Search Space
- Results of the hyperparameter optimization
- Analysis of the different RL algorithms
- Analysis of the different state spaces
- Analysis of the network parameters
- Analysis of the training parameters
- Statistical analysis of the results
- Performance of the controllers
- Conclusion and Outlook
- Appendix
- Results of the hyperparameter optimization
- Statistical analysis of the hyperparameter optimization results
- Step response of the controllers
- Lists