Deep Learning Fusion Model Based Electricity Load Forecasting for Extreme Weather
DOI:
https://doi.org/10.54097/njhzax84Keywords:
Deep learning; Extreme weather; Power load forecasting; Long and short-term memory network; XGBoost.Abstract
Aiming at the problem of insufficient accuracy of power load forecasting in extreme weather, a spatio-temporally aligned database is constructed based on the multi-source data (including 15-minute-level load profiles, 0.25°×0.25° gridded meteorological data, and EM-DAT standard disaster records) for the provincial power grids in the years 2015-2023. Fifteen key indicators are screened out by the random forest algorithm, and the hybrid LSTM-XGBoost model is innovatively proposed and dynamically weighted and optimised by Lasso regression. Tests show that the RMSE of the model is controlled in the range of 2.49~3.16 under typhoon weather, and the accuracy is improved by more than 45% compared with a single model. In the extreme high temperature event in North China in 2023, the prediction error of the model 6 hours in advance is only 4.7%. The technique has been validated in actual grid operation, providing reliable decision support for extreme weather power dispatch with important engineering application value.
Downloads
References
[1] Pang Hao, Gao Jinfeng, Du Yaoheng. Short-term load probability density prediction method based on time convolution network quantile regression[J]. Grid Technology, 2020, 44(04):1343-1350.
[2] Yuanyuan Wang, Jun Chen, Xiaoqiao Chen, Xiangjun Zeng, etal. Short-Term Load Forecasting for Industrial Customers based on TCN-LightGBM[J]. IEEE Transactions on Power Systems, 2021, 36(3):1984-1997. DOI: https://doi.org/10.1109/TPWRS.2020.3028133
[3] Luo Zhao, Wu Yuhou, Zhu Jiaxiang, et al. Ultra-short-term power prediction of wind power based on multi-scale time series block self-coding Transformer neural network model[J]. Grid Technology, 2023, 47(09): 3527-3537.
[4] Li Lianbing, Gao Guoqiang, Wu Weqiang, et al. Short-term day-ahead prediction of wind power considering feature recombination and improved transformer[J]. Grid Technology, 2024, 48(04):
[5] Chen ZY, Liu JB, Li C, et al. Ultra-short-term power load forecasting based on combined LSTM and XGBoost model[J]. Grid Technology, 2020, 44(02): 614-620.
[6] Liu Yahun, Zhao Qian. Ultra-short-term power load forecasting by CNN-LSTM based on clustered empirical modal decomposition[J]. Power Grid Technology, 2021, 45(11):4444-4451.
[7] Cheng Tong, Linghua Zhang, Hao Li, Yin Ding. temporal inception convolutional network based on multi-head attention for ultra-short-term load forecasting.IET Gener.Transm.Distrib[J], 2022, 16:1680-1696. DOI: https://doi.org/10.1049/gtd2.12394
[8] Meng Heng, Zhang T, Wang J, et al. A multi-node short-term power load forecasting method based on the fusion of multi-scale spatio-temporal graph convolutional network and Transformer [J/OL]. Power Grid Technology, 2024:1-14.
[9] Hewamalage H, Bergmeir C, Bandara K. Recurrent Neural Networks for Time Series Forecasting: Current status and future directions[J]. International Journal of Forecasting, 2021, 37(1):388-427. DOI: https://doi.org/10.1016/j.ijforecast.2020.06.008
[10] Kong W, Dong Z Y, Jia Y. Short-Term Residential Load Forecasting Based on LSTM Recurrent Neural Network [J]. IEEE Transactions on Smart Grid, 2019, 10 (1):841-851. DOI: https://doi.org/10.1109/TSG.2017.2753802
Downloads
Published
Issue
Section
License
Copyright (c) 2025 Highlights in Science, Engineering and Technology
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
