.png){kind=icon}
.png){kind=icon}
Return
LLM-Empowered Decision-Focused Learning for the Operation of Local Energy Communities
.png){kind=icon}
PRE
AI
Y
Y
D
Y
DOI：10.1109/TSG.2026.3678781.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Energy forecasting and dispatch are two essential procedures of local energy community (LEC) operations. In practice, these two procedures are often modeled and solved independently, leading to misaligned objectives, i.e., minimizing forecasting errors and costs. Recently, decision-focused learning (DFL) has been proposed to integrate these two procedures, but most existing works can merely be applied to a specific forecasting model or optimization problem. These methods are also technically complex for the operators to understand and implement. Therefore, this work proposes a large language model (LLM)-empowered DFL framework. Three LLM-based agents collaborate to fine-tune forecasts: one performs few-shot learning, one analyzes strategies, and one handles reflection. The case studies are conducted across four settings, two of which are challenging for existing methods, and tested with LLMs from three different providers. The results show that our method can achieve lower costs than existing approaches and requires no additional model training.
Keywords:
Decision-focused learning
large language model
load forecasting
decision-making
Journal
IF:
9.8
Papers:
5.6K
Citations:
4.3W
Organization
Cited Papers