Adaptive Stabilizing Control of Smart Transformer Based on Reinforcement Learning Optimization

The stability of power systems is currently facing significant challenges due to the increasing penetration of distributed energy resources (DERs). To improve stability, the smart transformer (ST) can be used to reshape the grid-side impedance that interacts with low-voltage (LV) side DER inverters by adjusting the control strategy. This paper first establishes the impedance models of the ST LV converter and DER inverter and analyzes the stability of the system. Specifically, the impact of the load and output power of the DER inverter on stability is investigated. A biquad filter-based active damping scheme is then adopted to improve system stability, with filter parameters pre-designed based on the root-locus curve. However, this design method is dependent on the system model and lacks adaptability, limiting the system stability region. To address this issue, this paper proposes using reinforcement learning (RL) algorithms to achieve adaptability in stabilizing control, as RL can achieve model-free training. The results show that the proposed approach can further broaden the region of stable operation for the system. Finally, simulations and experiments are used to demonstrate the effectiveness of adaptive stabilizing control.