The study introduces M2WLLM, a model that utilizes Large Language Models (LLMs) for ultra-short-term wind power forecasting.
M2WLLM combines textual information and numerical data to improve wind power forecasting accuracy through multi-modal data integration.
The model includes a Prompt Embedder and a Data Embedder to effectively fuse prompts and numerical inputs within the LLMs framework.
Empirical evaluations on wind farm data from Chinese provinces show that M2WLLM outperforms existing methods like GPT4TS in accuracy and few-shot learning capabilities.