Xiangkun Li

Xiangkun Li, Theodor Borsche, Göran Andersson

Increased distributed Photo-Voltaic (PV) generation leads to an increase in voltages and unwarranted backflows into the grid. This paper investigates Demand Response (DR) with Electric Water Heaters (EWHs) as a way to increase the PV hosting capacity of a low-voltage feeder. A control strategy relying only on power measurements at the transformer is proposed. Flexible loads are optimally dispatched considering energy acquisition costs, a PV shedding penalty, and power and energy constraints. Furthermore, grouping of loads and PV plants is investigated, and switching penalties are used to reduce the unnecessary switching of loads. It is shown that this strategy can substantially increase the PV hosting capacity of a Low-Voltage (LV) feeder, even when only basic controllability is available.

Xunlan Zhou, Xuanlin Chen, Shaowei Zhang et al.

Designing dense reward functions is pivotal for efficient robotic Reinforcement Learning (RL). However, most dense rewards rely on manual engineering, which fundamentally limits the scalability and automation of reinforcement learning. While Vision-Language Models (VLMs) offer a promising path to reward design, naive VLM rewards often misalign with task progress, struggle with spatial grounding, and show limited understanding of task semantics. To address these issues, we propose MARVL-Multi-stAge guidance for Robotic manipulation via Vision-Language models. MARVL fine-tunes a VLM for spatial and semantic consistency and decomposes tasks into multi-stage subtasks with task direction projection for trajectory sensitivity. Empirically, MARVL significantly outperforms existing VLM-reward methods on the Meta-World benchmark, demonstrating superior sample efficiency and robustness on sparse-reward manipulation tasks.