Safety-Guaranteed Voltage Regulation of Distribution Networks Using A Learning-Assisted Switching Tube Model Predictive Control Approach

This paper proposes a learning-assisted switching tube model predictive control (S-TMPC) approach to achieve absolutely safety-guaranteed voltage regulation in distribution networks (DNs). In order to model time-varying operation modes, the switching dynamics have been introduced based on respective equilibrium points in DNs. Then, to quantify the safe operation region of each mode, an active learning algorithm with a reliable oracle is developed to estimate the safety-guaranteed reachable set, which simplifies the conventionally linear-matrix-inequalitiesbased (LMI-based) computation. Next, considering power disturbances of renewable energy generation, a tube model predictive control approach for safe voltage regulation is introduced, and its exponential stability condition is given. Once the voltage deviation exceeds the boundary of the operation mode, a learning-assisted switching mechanism, of which the switching action is enabled by admissible minimum dwell-time analysis of switches among safe reachable sets (i.e., pre-learned tubes) is developed to absolutely ensure a safe voltage regulation. Finally, the effectiveness of the proposed voltage regulation method is verified by representative simulations and experiment tests.