Cross-Modal Knowledge Alignment for Power Load Forecasting

Accurate power load forecasting is vital for renewable energy integration. Existing deep learning-based load forecasting models primarily focus on numerical patterns but lack the ability to incorporate external semantic knowledge, limiting their adaptability to dynamic energy environments. To address these challenges, this paper proposes a novel Cross-Modal Knowledge Alignment (CMKA) method, which utilizes Large Language Models (LLMs) to obtain accurate power load forecasting. The proposed method can efficiently map time series data to the latent representation space of textual data and achieve deep matching between time series and natural language modalities. The method can activate the LLM's ability to understand and reason about time series data. The proposed model achieves sound performance in power load forecasting.