Zero-shot load forecasting for integrated energy systems: A large language model-based framework with multi-task learning

The growing penetration of renewable energy sources in power systems has increased the complexity and uncertainty of load forecasting, especially for integrated energy systems with multiple energy carriers. Traditional forecasting methods heavily rely on historical data and demonstrate limited transferability across different scenarios, creating significant challenges for emerging applications in smart grids and energy internet. This paper proposes the TSLLM-Load Forecasting Mechanism, a novel zero-shot load forecasting framework based on large language models (LLMs) to address these challenges. The framework consists of three key components: a data preprocessing module that handles multi-source energy load data, a time series prompt generation module that bridges the semantic gap between energy data and LLMs through multi-task learning and similarity alignment, and a prediction module that leverages pre-trained LLMs for accurate forecasting. We validate the framework's effectiveness on real-world datasets comprising load profiles from Australian solar-powered households. In conventional testing, our method achieves competitive performance, ranking third among nine methods. In zero-shot prediction, our framework demonstrates 10.8 % MSE improvement and 12.5 % MAE improvement compared to Informer, ranking second overall after TimeMixer. Most significantly, few-shot learning experiments reveal exceptional capabilities under extreme data constraints, with our method achieving optimal performance when trained with only 1 % of available data, representing 40.8 % MSE improvement compared to the best conventional method and 78.9 % improvement compared to existing LLM-based approaches. Large-scale transferability analysis shows our model outperforms baselines for 82 % of households when trained with minimal data from a single household. These results demonstrate the framework's potential for accurate and transferable load forecasting in integrated energy systems, particularly beneficial for renewable energy integration and smart grid applications where historical data availability is limited.