Journal of South China University of Technology(Natural Science Edition) ›› 2026, Vol. 54 ›› Issue (1): 42-52.doi: 10.12141/j.issn.1000-565X.250024

• Power & Electrical Engineering • Previous Articles     Next Articles

Short-Term Power Load Multi-Step Forecasting for Commercial Building Based on Improved Informer

ZHOU Xuan1,2,3   LI Kexin1   GUO Zixuan4   Yu Zhuliang2   YAN Junwei1,2,3   CAI Panpan1   

  1. 1. School of  Mechanical and Automotive Engineering, South China University of Technology, Guangzhou 510640, Guangdong, China; 2. Guangzhou Institute of Modern Industrial Technology, Guangzhou 511458, Guangdong, China; 3. Artificial Intelligence and Digital Economy Guangdong Province Laboratory(Guangzhou), Guangzhou 511442, Guangdong, China;

    4. School of Automation Science and Engineering, South China University of Technology, Guangzhou 510640, Guangdong, China

  • Online:2026-01-25 Published:2025-04-25

Abstract:

Short-term power load multi-step forecasting for commercial buildings plays a pivotal role in urban orderly power consumption and virtual power plant scheduling. The power load time series in commercial buildings is characterized by strong stochasticity, non-stationarity, and nonlinearity, and traditional iterative multi-step power load forecasting strategy suffer from error accumulation effects that degrade prediction accuracy., a short-term power load multi-step forecasting method based on Frequency Enhanced Channel Attention Mechanism (FECAM) -Sparrow Search Algorithm (SSA) -Informer is proposed. Based on the time-domain features output by the Informer encoder, the method use FECAM to adaptively model the frequency dependence between feature channels, further extracting the frequency-domain features of multi-dimensional input sequences. The decoder then integrates both time-frequency domain information to directly generate future multi-step load sequences. Furthermore, due to the lack of theoretical basis for the improved Informer hyperparameter settings, the SSA algorithm is used to optimize model hyperparameters such as learning rate, batch size, fully connected dimensions, and dropout rate. Experimental validation using annual load data from a commercial building in Guangzhou demonstrates that, compared with other deep learning models, the proposed model significantly improved prediction accuracy across varying forecast horizons (48-step, 96-step, 288-step, 480-step, 672-step), exhibiting superior performance in short-term power load multi-step forecasting.

Key words: power load forecasting for commercial buildings, frequency enhanced channel attention mechanism, Informer, sparrow search algorithm