Application of Large Language Models in Intelligent Preprocessing and Forecasting of Electricity Price

With the gradual liberalization of global energy markets, accurate electricity price forecasting is crucial for effective system operation and market-oriented reforms—especially for emerging Virtual Power Plants (VPPs) that rely on precise market insights. Traditional forecasting methods often struggle to account for the influence of complex unstructured data, such as news events, on price dynamics. To address this challenge, we propose a novel data preprocessing method that leverages large language models (LLMs) to automatically extract, classify, and derive insights from unstructured news articles (sourced from platforms such as WattClarity), which are then integrated with high-frequency electricity price data. To ensure data consistency, a weighted mean algorithm aggregates price records from 5-minute to 30-minute intervals. Our case study on a multi-source dataset demonstrates that this LLM-based technique not only effectively captures critical market signals from textual data—providing a robust framework for supporting market-driven decision-making for VPPs-but also overcomes the limitations of traditional forecasting approaches by harnessing the unique capabilities of LLMs. The integration of unstructured data processing with precise data aggregation underscores the practical advantages of our approach in modern electricity market applications.