Skip to content
DraconiticResearch
Search Sources API
178,033 studies31 sourcesUpdated 9 min ago
Peer-reviewedCC BYPLOS One

Long-term electricity load forecasting model based on multi-period and multi-scale analysis

  • Xiaolin Zhang1
  • Zhi Liu1
  • Chunyang Wang2
  • Zhen Qian3,4
4 affiliations
  1. 1School of Electronic Information Engineering, Changchun University of Science and Technology, Changchun, Jilin, China.
  2. 2Yahe School of Design and Engineering, Haikou University of Economics, Haikou, Hainan, China.
  3. 3Tsinghua University, Beijing, China.
  4. 4Beijing Jiaotong University, Beijing, China.
Read full textPDFDOICite

Abstract

Existing long-term power load forecasting models face significant challenges due to the complexity of non-stationary electricity data, inadequate information representation, and suboptimal prediction accuracy. To address these issues, this study introduces PowerNet, a novel forecasting framework based on a multi-period and multi-scale approach. PowerNet first employs Fast Fourier Transform (FFT) to perform multi-period analysis, transforming one-dimensional time series data into two-dimensional (2D) tensors to capture latent periodic patterns. A multi-scale information representation module is then implemented using computer vision-inspired techniques, including Convolutional Neural Networks (CNN), Gated Recurrent Units (GRU), and non-local attention mechanisms to extract robust features from these tensors. Finally, a fully connected layer performs feature reconstruction, while residual connections facilitate multi-step load forecasting. PowerNet was evaluated against state-of-the-art models such as Informer, LogTrans, and ETSformer across two public electricity datasets and one weather dataset. At prediction horizons of 96, 192, 336, and 720 time steps, experimental results demonstrate that PowerNet achieves competitive or superior accuracy on the evaluated datasets. Notably, compared with Autoformer, PowerNet reduces the Mean Absolute Error (MAE) by 18.77% to 37.41% and the Mean Square Error (MSE) by 18.23% to 47.74%, significantly enhancing the reliability of long-term electricity load forecasting.

Related

  1. Bi-Mamba-Load: A bidirectional structured state space model with sequence relationship adaptive perception for electricity load forecasting
    Junwei Zhu, Lingzhi Xi, Chongyang Liao +3
    PLOS OnePeer-reviewedCC BY 4.0NeuroscienceAtmospheric science & meteorologyPLOS
  2. OmicsBench: a long-horizon multi-turn benchmark of AI agents on research-study-scale bioinformatics
    Gideon Shaked, Siddharth Nagisetty
    arXivPreprintCC BY 4.0Full textBioinformatics & computational biologyGenetics & genomicsarXiv
  3. Leveraging Multi-Isotope Systems to Investigate Conflict, Food Insecurity, and Elevated Nitrogen Isotope Values in the Nazca Highlands of Peru During the Late Intermediate Period (1000-1450 CE).
    Sarah J Noe, Bridgette Degnan, Emily C Kracht +1
    American journal of biological anthropologyPeer-reviewedCC BYFull textEvolutionary biologyEurope PMCPMC
  4. Learning and forecasting shared evolutionary pathways to multi-drug resistance across global pathogens
    O. Aga, S. Moyo, J. Ferno +6
    bioRxivPreprintCC BYFull textEvolutionary biologybioRxiv
  5. Nation-scale recovery of complete environmental plasmids pushes the boundaries of plasmid diversity through long-read sequencing
    B. Sorensen, C. L. Brown, M. Sereika +4
    bioRxivPreprintCC BY-NCFull textMicrobiologybioRxiv

Cite

BibTeXRIS
BibTeX
@article{zhang2026longterm,
  title = {{Long-term electricity load forecasting model based on multi-period and multi-scale analysis}},
  author = {Xiaolin Zhang and Zhi Liu and Chunyang Wang and Zhen Qian},
  journal = {PLOS One},
  year = {2026},
  month = oct,
  doi = {10.1371/journal.pone.0361106},
  url = {https://doi.org/10.1371/journal.pone.0361106},
  note = {CC BY},
}
RIS
TY  - JOUR
TI  - Long-term electricity load forecasting model based on multi-period and multi-scale analysis
AU  - Xiaolin Zhang
AU  - Zhi Liu
AU  - Chunyang Wang
AU  - Zhen Qian
PY  - 2026
DA  - 2026/10/30
JO  - PLOS One
DO  - 10.1371/journal.pone.0361106
UR  - https://doi.org/10.1371/journal.pone.0361106
AB  - Existing long-term power load forecasting models face significant challenges due to the complexity of non-stationary electricity data, inadequate information representation, and suboptimal prediction accuracy. To address these issues, this study introduces PowerNet, a novel forecasting framework based on a multi-period and multi-scale approach. PowerNet first employs Fast Fourier Transform (FFT) to perform multi-period analysis, transforming one-dimensional time series data into two-dimensional (2D) tensors to capture latent periodic patterns. A multi-scale information representation module is then implemented using computer vision-inspired techniques, including Convolutional Neural Networks (CNN), Gated Recurrent Units (GRU), and non-local attention mechanisms to extract robust features from these tensors. Finally, a fully connected layer performs feature reconstruction, while residual connections facilitate multi-step load forecasting. PowerNet was evaluated against state-of-the-art models such as Informer, LogTrans, and ETSformer across two public electricity datasets and one weather dataset. At prediction horizons of 96, 192, 336, and 720 time steps, experimental results demonstrate that PowerNet achieves competitive or superior accuracy on the evaluated datasets. Notably, compared with Autoformer, PowerNet reduces the Mean Absolute Error (MAE) by 18.77% to 37.41% and the Mean Square Error (MSE) by 18.23% to 47.74%, significantly enhancing the reliability of long-term electricity load forecasting.
SN  - 1932-6203
ER  -