Informer-Based Long-Horizon Power Load Forecasting: An Empirical Study on the SG-HL Dataset

Long-horizon load forecasting is vital for dispatch and risk control. This study provides an engineering evaluation of Informer on a SmartGrid-Hourly Load (SG-HL) dataset. A unified protocol is adopted for 24h, 480h, and 720h tasks, with consistent preprocessing, rolling-window construction, and metrics. Under identical settings, Informer is compared with LSTM, GRU, and TCN. Results show trend-level consistency over long horizons with bounded errors; relative to baselines, accuracy is competitive and error growth is slower, while peak-load extremes remain challenging. The study summarizes practical configurations and offers a reproducible baseline for medium- and long-term scheduling and risk management.

• Publication year

  2025

• Venue

  2025 International Conference on Low Carbon and Smart Energy (ICLCSE)

• Publication date

  2025-10-24

• Fields of study

  Not labeled

• Identifiers
  DOI 10.1109/ICLCSE66889.2025.11389692
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Deep Reinforcement Learning: A Survey

  2022cited by this paper
• Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting

  2021cited by this paper
• Temporal Convolutional Networks Applied to Energy-Related Time Series Forecasting

  2020cited by this paper
• Time-series forecasting with deep learning: a survey

  2020cited by this paper
• Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting

  2020cited by this paper
• An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling

  2018cited by this paper
• Attention is All you Need

  2017cited by this paper
• Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation

  2014cited by this paper
• Long Short-Term Memory

  1997cited by this paper

• No citing papers are available for this paper.