Safe reinforcement learning-based control of voltage-forming grid inverters / Daniel Weber ; First Reviewer: Prof. Dr.-Ing. Joachim Böcker, Second Reviewer: Prof. Dr.-Ing. Oliver Wallscheid. Paderborn, 2026
Inhalt
- Acknowledgment
- Abstract
- Zusammenfassung
- Own Publications
- Nomenclature
- Contents
- Introduction
- Fundamentals
- Optimal Control for Voltage-Forming Inverters
- Voltage Source Inverter
- Reinforcement Learning Fundamentals
- Safeguarding in Continuous State and Action Space
- Derivation of the Relevant Constraints
- Example: Double Integrator
- Differential Constraints
- System Identification
- Power Sharing
- RL-Based Control of Voltage-Forming Grid Inverters
- Experimental Setup
- Digital Control Delay Compensation
- Reward Design
- Featurization
- Training via Edge Computing
- Hyperparameter Optimization in Real-Time Application
- Application: Constant, Resistive Load
- State-Space Representation of the Plant
- Safeguard in Real Time
- Data-Driven Safeguard
- Hyperparameter Configuration
- Training Behavior
- Steady-State Behavior
- Transient Behavior
- Black-Start Capability
- Operation with Active Output Constraints
- Computing Time Evaluation
- Unbalanced Loads
- Application: Arbitrary Loads
- Decentralized Inverter-Based Microgrid Application
- Conclusion and Outlook
- Appendix
- Deep Deterministic Policy Gradient Algorithm - Basics
- Hyperparameter Settings of the Used DDPG Agents
- Predictive Dead-Beat Controller
- Artificial Intelligence use Declaration
- Lists
- References