|
|
|
| Ultra-Short-Term Load Forecasting Based on Dual Decomposition Strategy |
| Li Nan1,2, Wang Hetong2, Zhang Guangyao3, Jin Chunxi4, Zhu Yabin5 |
1. Key Laboratory of Modern Power System Simulation and Control & Renewable Energy Technology Ministry of Education Northeast Electric Power University Jilin 132012 China;
2. School of Electrical Engineering Northeast Electric Power University Jilin 132012 China;
3. State Grid Shandong Electric Extrahigh Voltage Company Jinan 250000 China;
4. Anhui Jianghuai Automobile Group Co. Ltd Hefei 230022 China;
5. State Grid Shaanxi Electric Power Company Limited Xianyang Power Supply Company Xianyang 712000 China |
|
|
|
Abstract High renewable energy penetration accentuates the non-stationarity, nonlinearity, and randomness of power system loads, challenging power grid transient stability and planning. Accurate ultra-short-term load forecasting (USTLF) is thereby indispensable for the transient stability analysis of power systems. This study proposes a new USTLF method that integrates a dual decomposition strategy for data preprocessing with a hybrid multi-scale temporal convolutional network and bidirectional long short-term memory (MSTCN-BiLSTM) model for prediction.
Firstly, a dual decomposition strategy integrating empirical wavelet transform-frequency domain stability (EWT-FreqStab) and seasonal trend dispersion reminder (STDR) is proposed. Specifically, the empirical wavelet transform (EWT) adaptively decomposes load data into frequency sub-bands, while frequency-domain stability weighting effectively suppresses non-stationary effects in high-frequency components. Secondly, STDR algorithm accurately captures dual daily-weekly periodic patterns and extracts fluctuation-associated dispersion components. These decomposed elements, when combined with strongly correlated meteorological data, form a multivariate dataset enriched with multi-scale input features. Finally, a MSTCN-BiLSTM architecture is designed: this architecture utilizes multi-scale temporal convolutional layers with different dilation rates for local and global feature extraction, integrates a multi-head attention mechanism to dynamically weight meaningful patterns, incorporates bidirectional long short-term memory (BiLSTM) units to enhance the modeling of long-term temporal dependencies, and ultimately accomplishes the prediction task.
This study presents simulation results derived from annual load data across domestic and international regions during the 2021-2022 period. The dataset is split into training and test sets with a ratio of 7:3. To fully validate the effectiveness of the proposed method, a comparative experimental framework is developed from four distinct perspectives, which includes the following aspects: (1) An analysis of the impact of input window length on model performance. (2) A performance comparison between the proposed model and other mainstream forecasting models. (3) Comparative experiments examining the same prediction models integrated with the dual decomposition strategy versus those adopting traditional decomposition methods. (4) A comparative analysis of the dual decomposition strategy combined with different prediction models.
The experimental results demonstrate that when the input window length is 4, the innovative dual decomposition strategy, in conjunction with the MSTCN-BiLSTM prediction model, achieves outstanding performance metrics. Specifically, on the domestic dataset, it yields a MAE of 19.975 MW, a RMSE of 27.099 MW, and a MAPE of 0.831%. For the international dataset, the corresponding metrics are 24.373 MW (MAE), 32.492 MW (RMSE), and 0.392% (MAPE).
The following conclusions can be drawn from the simulation analysis: (1) The STDR decomposition algorithm performs multi-scale decomposition on load data, generating four-dimensional temporal-scale components that contain discrete fluctuation information characterizing periodic patterns. Building on this, the EWT-FreqStab algorithm further decomposes the load data into stable and effective multi-dimensional time-frequency components. Subsequently, these components are integrated with meteorological data of high correlation to construct a multivariate component matrix, thereby achieving comprehensive characterization of the multi-dimensional information inherent in the load data. (2) The proposed MSTCN-BiLSTM network effectively integrates feature information across diverse time scales through a series-parallel TCN architecture. The embedded multi-head attention mechanism strengthens the model's capacity to capture critical features, while the bidirectional temporal modeling capability of BiLSTM establishes a feature extraction framework that balances short-term details and long-term dependencies—thereby offering an effective technical pathway for enhancing the accuracy of ultra-short-term load forecasting.
|
|
Received: 30 July 2025
|
|
|
|
|
|
[1] 孙秋野, 刘广亮, 王一帆. 能源互联网中能源终端的研究综述及展望[J]. 电网技术, 2025, 49(5): 1792-1805.
Sun Qiuye, Liu Guangliang, Wang Yifan.Review and prospect of energy terminal in energy Internet[J]. Power System Technology, 2025, 49(5): 1792-1805.
[2] 刘晓军, 熊健, 王艺博, 等. 考虑不确定变量变分模态分解及绿证-碳联合交易的综合能源系统经济优化调度[J]. 电工技术学报, 2025, 40(13): 4276-4291.
Liu Xiaojun, Xiong Jian, Wang Yibo, et al.Economic optimization of integrated energy system scheduling considering uncertainty variables variational mode decomposition and green certificate-carbon joint trading[J]. Transactions of China Electrotechnical Society, 2025, 40(13): 4276-4291.
[3] 韩富佳, 王晓辉, 乔骥, 等. 基于人工智能技术的新型电力系统负荷预测研究综述[J]. 中国电机工程学报, 2023, 43(22): 8569-8592.
Han Fujia, Wang Xiaohui, Qiao Ji, et al.Review on artificial intelligence based load forecasting research for the new-type power system[J]. Proceedings of the CSEE, 2023, 43(22): 8569-8592.
[4] 郇嘉嘉, 李代猛, 杜云飞, 等. 基于Prophet算法和Blending集成学习的实时负荷中期预测[J]. 电力自动化设备, 2024, 44(4): 178-183.
Huan Jiajia, Li Daimeng, Du Yunfei, et al.Mid-term forecasting of real-time load based on Prophet algorithm and Blending integrated learning[J]. Electric Power Automation Equipment, 2024, 44(4): 178-183.
[5] 王凌云, 周翔, 田恬, 等. 基于多维气象信息时空融合和MPA-VMD的短期电力负荷组合预测模型[J]. 电力自动化设备, 2024, 44(2): 190-197.
Wang Lingyun, Zhou Xiang, Tian Tian, et al.Combination forecasting model of short-term power load based on multi-dimensional meteorological information spatio-temporal fusion and MPA-VMD[J]. Electric Power Automation Equipment, 2024, 44(2): 190-197.
[6] 张宇帆, 艾芊, 林琳, 等. 基于深度长短时记忆网络的区域级超短期负荷预测方法[J]. 电网技术, 2019, 43(6): 1884-1892.
Zhang Yufan, Ai Qian, Lin Lin, et al.A very short-term load forecasting method based on deep LSTM RNN at zone level[J]. Power System Technology, 2019, 43(6): 1884-1892.
[7] 李云松, 张智晟. 考虑综合需求响应的Transformer-图神经网络综合能源系统多元负荷短期预测[J]. 电工技术学报, 2024, 39(19): 6119-6128.
Li Yunsong, Zhang Zhisheng.Transformer based multi load short-term forecasting of integrated energy system considering integrated demand response[J]. Transactions of China Electrotechnical Society, 2024, 39(19): 6119-6128.
[8] 梁露, 张智晟. 基于多尺度特征增强DHTCN的电力系统短期负荷预测研究[J]. 电力系统保护与控制, 2023, 51(10): 172-179.
Liang Lu, Zhang Zhisheng.Short-term load forecasting of a power system based on multi-scale feature enhanced DHTCN[J]. Power System Protection and Control, 2023, 51(10): 172-179.
[9] 陈纬楠, 胡志坚, 岳菁鹏, 等. 基于长短期记忆网络和LightGBM组合模型的短期负荷预测[J]. 电力系统自动化, 2021, 45(4): 91-97.
Chen Weinan, Hu Zhijian, Yue Jingpeng, et al.Short-term load prediction based on combined model of long short-term memory network and light gradient boosting machine[J]. Automation of Electric Power Systems, 2021, 45(4): 91-97.
[10] 周洲, 焦文玲, 任乐梅, 等. 蚁群算法分配权重的燃气日负荷组合预测模型[J]. 哈尔滨工业大学学报, 2021, 53(6): 177-183.
Zhou Zhou, Jiao Wenling, Ren Lemei, et al.Combined forecasting model of gas daily load based on weight distribution of ant colony algorithm[J]. Journal of Harbin Institute of Technology, 2021, 53(6): 177-183.
[11] 朱卫涛, 邹文文, 贾钦, 等. 基于DWT-SOM-HFS的配电台区短期负荷预测研究与应用[J]. 智慧电力, 2023, 51(6): 78-85.
Zhu Weitao, Zou Wenwen, Jia Qin, et al.Research and application of short term load forecasting in distribution station area based on DWT-SOM-HFS[J]. Smart Power, 2023, 51(6): 78-85.
[12] 钟吴君, 李培强, 涂春鸣. 基于EEMD-CBAM-BiLSTM的牵引负荷超短期预测[J]. 电工技术学报, 2024, 39(21): 6850-6864.
Zhong Wujun, Li Peiqiang, Tu Chunming.Ultra-short-term prediction of traction load based on EEMD-CBAM-BiLSTM[J]. Transactions of China Electrotechnical Society, 2024, 39(21): 6850-6864.
[13] Cai Changchun, Li Yuanjia, Su Zhenghua, et al.Short-term electrical load forecasting based on VMD and GRU-TCN hybrid network[J]. Applied Sciences, 2022, 12(13): 6647.
[14] Gilles J.Empirical wavelet transform[J]. IEEE Transactions on Signal Processing, 2013, 61(16): 3999-4010.
[15] Aja-Fernández S, Alberola-López C.On the estimation of the coefficient of variation for anisotropic diffusion speckle filtering[J]. IEEE Transactions on Image Processing, 2006, 15(9): 2694-2701.
[16] Dudek G.STD: a seasonal-trend-dispersion decomposition of time series[J]. IEEE Transactions on Knowledge and Data Engineering, 2023, 35(10): 10339-10350.
[17] 刘巍炜, 周羽生, 周文晴, 等. 考虑异方差性的城市电网电动汽车充电负荷预测[J]. 电力系统自动化, 2024, 48(15): 54-63.
Liu Weiwei, Zhou Yusheng, Zhou Wenqing, et al.Charging load forecasting for electric vehicles in urban power grid considering heteroscedasticity[J]. Automation of Electric Power Systems, 2024, 48(15): 54-63.
[18] 李楠, 姜涛, 隋想, 等. 一种时频尺度下的多元短期电力负荷组合预测方法[J]. 电力系统保护与控制, 2024, 52(13): 47-58.
Li Nan, Jiang Tao, Sui Xiang, et al.A multi-component short-term power load combination forecasting method on a time-frequency scale[J]. Power System Protection and Control, 2024, 52(13): 47-58.
[19] 时培明, 郭轩宇, 杜清灿, 等. 基于TCN-BiLSTM-Attention-ESN的光伏功率预测[J]. 太阳能学报, 2024, 45(9): 304-316.
Shi Peiming, Guo Xuanyu, Du Qingcan, et al.Photovoltaic power prediction based on TCN-BiLSTM-attention-ESN[J]. Acta Energiae Solaris Sinica, 2024, 45(9): 304-316.
[20] 李延珍, 王海鑫, 杨子豪, 等. 基于非侵入式负荷分解的家庭负荷两阶段超短期负荷预测模型[J]. 电工技术学报, 2024, 39(11): 3379-3391.
Li Yanzhen, Wang Haixin, Yang Zihao, et al.Two-stage ultra-short-term load forecasting model of household appliances based on non-intrusive load disaggregation[J]. Transactions of China Electro-technical Society, 2024, 39(11): 3379-3391.
[21] 周思思, 李勇, 郭钇秀, 等. 考虑时序特征提取与双重注意力融合的TCN超短期负荷预测[J]. 电力系统自动化, 2023, 47(18): 193-205.
Zhou Sisi, Li Yong, Guo Yixiu, et al.Ultra-short-term load forecasting based on temporal convolutional network considering temporal feature extraction and dual attention fusion[J]. Automation of Electric Power Systems, 2023, 47(18): 193-205.
[22] 王光华, 张纪欣, 崔良, 等. 基于双重注意力变换模型的分布式屋顶光伏变电站级日前功率预测[J]. 全球能源互联网, 2024, 7(4): 393-405.
Wang Guanghua, Zhang Jixin, Cui Liang, et al.Substation-level distributed rooftop photovoltaic power day-ahead prediction based on double attention mechanism transformer model[J]. Journal of Global Energy Interconnection, 2024, 7(4): 393-405.
[23] Lee M.Mathematical analysis and performance evaluation of the GELU activation function in deep learning[J]. Journal of Mathematics, 2023, 2023(1): 4229924.
[24] 罗澍忻, 麻敏华, 蒋林, 等. 考虑多时间尺度数据的中长期负荷预测方法[J]. 中国电机工程学报, 2020, 40(增刊1): 11-19.
Luo Shuxin, Ma Minhua, Jiang Lin, et al.Medium and long-term load forecasting method considering multi-time scale data[J]. Proceedings of the CSEE, 2020, 40(S1): 11-19.
[25] 何安明, 赵鑫, 吴立刚, 等. 基于双向长短期记忆网络的区域电网新能源消纳预测算法[J]. 电气技术, 2023, 24(3): 23-30.
He Anming, Zhao Xin, Wu Ligang, et al.Prediction algorithm of new energy consumption in regional power grid based on bidirectional long short term memory network[J]. Electrical Engineering, 2023, 24(3): 23-30.
[26] 杜伟, 王圣, 李健, 等. 基于CNN-LSTM-AM模型的储能锂离子电池荷电状态预测[J]. 电工技术学报, 2025, 40(9): 2982-2995.
Du Wei, Wang Sheng, Li Jian, et al.Prediction of state of charge for energy storage lithium-ion batteries based on CNN-LSTM-AM model[J]. Transactions of China Electrotechnical Society, 2025, 40(9): 2982-2995.
[27] 任建吉, 位慧慧, 邹卓霖, 等. 基于CNN-BiLSTM-Attention的超短期电力负荷预测[J]. 电力系统保护与控制, 2022, 50(8): 108-116.
Ren Jianji, Wei Huihui, Zou Zhuolin, et al.Ultra-short-term power load forecasting based on CNN-BiLSTM-Attention[J]. Power System Protection and Control, 2022, 50(8): 108-116.
[28] Chen Zihan, Chen Guici, Wang Wenbo, et al.Short-term power load forecasting based on CEEMDAN-CNN-LSTM hybrid modeling[C]//2024 12th International Conference on Intelligent Control and Information Processing (ICICIP), Nanjing, China, 2024: 9-16.
[29] Li Siting, Cai Huafeng.Short-term power load forecasting using a VMD-crossformer model[J]. Energies, 2024, 17(11): 2773.
[30] 陈海鹏, 李赫, 阚天洋, 等. 考虑风电时序特性的深度小波-时序卷积网络超短期风功率预测[J]. 电网技术, 2023, 47(4): 1653-1665.
Chen Haipeng, Li He, Kan Tianyang, et al.DWT-DTCNA ultra-short-term wind power prediction considering wind power timing characteristics[J]. Power System Technology, 2023, 47(4): 1653-1665. |
|
|
|
2026, Vol. 41 
