Suzhou Electric Appliance Research Institute
期刊号: CN32-1800/TM| ISSN1007-3175

Article retrieval

文章检索

首页 >> 文章检索 >> 最新索引

基于注意力机制的ABG-GCA模型中长期风电功率预测

来源:电工电气发布时间:2025-03-03 15:03 浏览次数:9

基于注意力机制的ABG-GCA模型中长期风电功率预测

蒲海涛1,2,代英健1
(1 山东科技大学 电气与自动化工程学院,山东 青岛 266590;
2 山东科技大学济南校区 电气信息系,山东 济南 250031)
 
    摘 要:风电功率预测对电力系统的稳定性和经济性具有重要意义。针对已有模型预测时间较长和预测精度存在较大误差的问题,提出了一种新型的 ABG-GCA 模型,该模型通过 Autoformer 的自相关机制与基于全局注意力机制的双向门控循环单元将处理好的数据进行并行预测,对各分量的预测值利用交叉注意力机制来进行权重分配形成高效准确功率的预测结果。实验结果表明,该模型在预测精度和时间效率方面优于传统模型,能够有效捕捉风电功率的变化趋势,对于不同季节的预测自适应性极强且预测精度高。
    关键词: 风电功率预测;二次分解技术;ABG-GCA 模型;中长期预测;自相关机制;全局注意力机制;交叉注意力机制;预测精度
    中图分类号:TM614     文献标识码:A     文章编号:1007-3175(2025)02-0010-09
 
Medium and Long Term Wind Power Prediction by ABG-GCA
Model Based on Attention Mechanism
 
PU Hai-tao1, 2, DAI Ying-jian1
(1 College of Electrical Engineering and Automation, Shandong University of Science and Technology, Qingdao 266590, China;
2 Electrical Information Department, Shandong University of Science and Technology-Jinan Campus, Jinan 250031, China)
 
    Abstract: The wind power prediction is of great significance to the stability and economy of power system. In this paper, a new ABGGCA model is proposed to solve the problem that the existing model has a long prediction time and a large error in prediction accuracy. The model uses the autocorrelation mechanism of Autoformer and the bidirectional gated recurrent unit based on the global attention mechanism to predict the processed data in parallel. The cross-attention mechanism is used to assign weights to the predicted values of each component to form an efficient and accurate power prediction result. The experimental results show that the model is superior to the traditional model,in terms of prediction accuracy and time efficiency which can effectively capture the change trend of wind power and prediction for different seasons has strong adaptability and high precision.
    Key words: wind power prediction; secondary decomposition technique; ABG-GCA model; medium and long term prediction; autocorrelation mechanism; global attention mechanism; cross-attention mechanism; prediction accuracy
 
参考文献
[1] 汪欣. 基于神经网络的风电功率优化预测方法[D]. 上海:上海交通大学,2020.
[2] ZHU Changsheng, ZHU Lina.Wind Speed Short-Term Prediction Based on Empirical Wavelet Transform, Recurrent Neural Network and Error Correction[J].Journal of Shanghai Jiaotong University(Science),2024,29(2) :297-308.
[3] CHEN Gonggui, LI Lijun, ZHANG Zhizhong, et al.Short-term wind speed forecasting with principle-subordinate predictor based on Conv-LSTM and improved BPNN[J].IEEE Access,2020,8 :67955-67973.
[4] 程杰,陈鼎,李春,等. 基于 GWO-CNN-BiLSTM 的超短期风电预测[J] . 科学技术与工程,2023,23(35) :15091-15099.
[5] 符杨,任子旭,魏书荣,等. 基于改进 LSTM-TCN 模型的海上风电超短期功率预测[J] . 中国电机工程学报,2022,42(12) :4292-4302.
[6] 郎伟明,麻向津,周博文,等.基于 LSTM 和非参数核密度估计的风电功率概率区间预测[J].智慧电力,2020,48(2) :31-37.
[7] WANG W, FENG B, HUANG G, et al.Conformal asymmetric multi-quantile generative transformer for day-ahead wind power interval prediction[J].Applied Energy,2023,333 :120634.
[8] BENTSEN L D, WARAKAGODA N D, STENBRO R, et al. pSatiotemporal wind speed forecasting using graph networks and novel transformer architectures[J].Applied Energy,2023,333 :120565.
[9] WANG Lei, HE Yigang, LI Lie, et al.A novel approach to ultra-short-term multi-step wind power predictions based on encoder-decoder architecture in natural language processing[J].Journal of Cleaner Production,2022,354 :131723.
[10] 骆钊,吴谕侯,朱家祥,等. 基于多尺度时间序列块自编码 Transformer 神经网络模型的风电超短期功率预测[J]. 电网技术,2023,47(9) :3527-3536.
[11] 林铮,刘可真,沈赋,等. 考虑海上风电多机组时空特性的超短期功率预测模型[J] . 电力系统自动化,2022,46(23) :59-66.
[12] WU Haixu, XU Jiehui, WANG Jianmin, et al.Autoformer: Decomposition transformers with autocorrelation for long-term series forecasting[J].Advances in Neural Information Processing Systems,2021,34 :22419-22430.
[13] 王渝红,史云翔,周旭,等. 基于时间模式注意力机制的 BiLSTM 多风电机组超短期功率预测[J]. 高电压技术,2022,48(5) :1884-1892.
[14] 宋柯. 基于多时间尺度及注意力机制的风电功率预测技术研究[D]. 重庆:重庆理工大学,2023.
[15] 李静茹,姚方. 引入注意力机制的 CNN 和 LSTM 复合风电预测模型[J]. 电气自动化,2022,44(6) :4-6.
[16] 王家乐,张耀,林帆,等. 基于自注意力特征提取的光伏功率组合概率预测[J] . 太阳能学报,2024,45(12) :123-131.
[17] CAI J , ZHANG K , JIANG H . Power Quality Disturbance Classification Based on Parallel Fusion of CNN and GRU [J] . Energies,2023,16(10) :4029.
[18] 乔石,王磊,张鹏超,等. 基于时间模式注意力机制的 GRU 短期负荷预测[J] . 电力系统及其自动化学报,2023,35(10) :49-58.
[19] 龙铖,余成波,何铖,等. 基于双重注意力机制 CNN-BiLSTM 与 LightGBM 误差修正的超短期风电功率预测[J].电气工程学报,2024,19(2) :138-145.
[20] LUONG M T, PHAM H, MANNING C D.Effective approaches to attention-based neural machine translation[J].ArXiv Preprint ArXiv,2015,1508 :04025.