基于损失函数改进和补丁时序Transformer网络的风功率超短期多步预测

晏吴宇歆; 张海波; 刘童蕙; 黄松涛; 尚国政

doi:10.19912/j.0254-0096.tynxb.2024-1903

PDF(2268 KB)

太阳能学报 ›› 2025, Vol. 46 ›› Issue (6) : 510-521. DOI: 10.19912/j.0254-0096.tynxb.2024-1903

基于损失函数改进和补丁时序Transformer网络的风功率超短期多步预测

晏吴宇歆¹, 张海波¹, 刘童蕙¹, 黄松涛², 尚国政²

作者信息 +

ULTRA-SHORT-TERM MULTI-STEP PREDICTION OF WIND POWER BASED ON LOSS FUNCTION IMPROVEMENT AND PATCH TIMING TRANSFORMER NETWORK

Yan Wuyuxin¹, Zhang Haibo¹, Liu Tonghui¹, Huang Songtao², Shang Guozheng²

Author information +

文章历史 +

摘要

为提高风电功率超短期多步预测的精度,提出一种基于损失函数改进和补丁（Patch）时序Transformer网络的风功率超短期多步预测模型。首先,通过风功率数据的图像异常检测与清洗算法进行数据预处理,提升数据质量;其次,为增强Transformer模型的鲁棒性并加强局部序列依赖的捕捉能力,在原始Transformer结构中引入补丁模块和通道独立策略。最后,为进一步过滤噪声并提高序列预测中的形状变化感知能力,设计一种新颖的多元非线性损失函数。实验结果表明,所提出的模型在多项误差指标上显著优于对比模型,有效提升了超短期风电功率的多步预测精度。

Abstract

To enhance the accuracy of ultra-short-term multi-step wind power forecasting, this study proposes a novel model that integrates improvements in the loss function with a patch-based temporal Transformer network. Specifically, an image-based anomaly detection and cleaning algorithm is firstly employed for data preprocessing, thereby enhancing the quality of the wind power data. Subsequently, to improve the robustness of the Transformer architecture and to strengthen its ability to capture local sequential dependencies, a patch module and a channel-independent strategy are incorporated into the standard Transformer framework. Finally, a novel multivariate nonlinear loss function is designed to effectively filter noise and to enhance the model's sensitivity to shape variations during sequence prediction. Extensive experimental results demonstrate that the proposed approach significantly outperforms several baseline models across multiple error metrics, thereby achieving substantial improvements in ultra-short-term multi-step wind power forecasting accuracy.

导出引用

晏吴宇歆, 张海波, 刘童蕙, 黄松涛, 尚国政. 基于损失函数改进和补丁时序Transformer网络的风功率超短期多步预测[J]. 太阳能学报. 2025, 46(6): 510-521 https://doi.org/10.19912/j.0254-0096.tynxb.2024-1903

Yan Wuyuxin, Zhang Haibo, Liu Tonghui, Huang Songtao, Shang Guozheng. ULTRA-SHORT-TERM MULTI-STEP PREDICTION OF WIND POWER BASED ON LOSS FUNCTION IMPROVEMENT AND PATCH TIMING TRANSFORMER NETWORK[J]. Acta Energiae Solaris Sinica. 2025, 46(6): 510-521 https://doi.org/10.19912/j.0254-0096.tynxb.2024-1903

中图分类号： TM731

参考文献

[1] 国家能源局发布2023年全国电力工业统计数据[J]. 电力科技与环保, 2024, 40(1): 95.
National Energy Administration releases 2023 national power industry statistics[J]. Electric power science and technology and environmental protection, 2024, 40(1): 95.
[2] 孙荣富, 张涛, 和青, 等. 风电功率预测关键技术及应用综述[J]. 高电压技术, 2021, 47(4): 1129-1143.
SUN R F, ZHANG T, HE Q, et al.Review on key technologies and applications in wind power forecasting[J]. High voltage engineering, 2021, 47(4): 1129-1143.
[3] 舒印彪, 张丽英, 张运洲, 等. 我国电力碳达峰、碳中和路径研究[J]. 中国工程科学, 2021, 23(6): 1-14.
SHU Y B, ZHANG L Y, ZHANG Y Z, et al.Research on the path of carbon peaking and carbon neutrality in, China's power industry[J]. China engineering science, 2021, 23(6): 1-14.
[4] 史加荣, 赵丹梦, 王琳华, 等. 基于RR-VMD-LSTM的短期风电功率预测[J]. 电力系统保护与控制, 2021, 49(21): 63-70.
SHI J R,ZHAO D M,WANG L H,et al.Short term wind power prediction based on RR-VMD-LSTM[J]. Power system protection and control, 2021, 49(21): 63-70.
[5] 赵泽妮, 云斯宁, 贾凌云, 等. 基于统计模型的短期风能预测方法研究进展[J]. 太阳能学报, 2022, 43(11): 224-234.
ZHAO Z N, YUN S N, JIA L Y, et al.Recent progress in short-term forecasting of wind energy based on statistical models[J]. Acta energiae solaris sinica, 2022, 43(11): 224-234.
[6] 兰昆, 吴战波, 赵泽妮, 等. 基于多因子的太阳辐照度预测方法研究进展[J]. 太阳能学报, 2024, 45(5): 593-601.
LAN K, WU Z B, ZHAO Z N, et al.Advance in prediction methods for solar irradiance based on multiple-factors[J]. Acta energiae solaris sinica, 2024, 45(5): 593-601.
[7] 贾凌云, 云斯宁, 赵泽妮, 等. 神经网络短期光伏发电预测的应用研究进展[J]. 太阳能学报, 2022, 43(12): 88-97.
JIA L Y, YUN S N, ZHAO Z N, et al.Recent progress of short-term forecasting of photovoltaic generation based on artificial neural networks[J]. Acta energiae solaris sinica, 2022, 43(12): 88-97.
[8] JIA L Y, YUN S N, ZHAO Z N, et al.Improving short-term forecasting of solar power generation by using an EEMD-BiGRU model: a comparative study based on seven standalone models and six hybrid models[J]. International journal of green energy, 2024, 21(14): 3135-3158.
[9] 刘长良, 赵陆阳, 王梓齐, 等. 基于时空注意力-Seq2Seq模型的多风电机组多步风速预测算法[J]. 太阳能学报, 2023, 44(8): 420-429.
LIU C L, ZHAO L Y, WANG Z Q, et al.Multi-step wind speed prediction algorithm of multiple wind turbines based on spatial-temporal attention-Seq2Seq model[J]. Acta energiae solaris sinica, 2023, 44(8): 420-429.
[10] 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.
[11] BENTSEN L Ø, WARAKAGODA N D, STENBRO R, et al.Spatio-temporal wind speed forecasting using graph networks and novel Transformer architectures[J]. Applied energy, 2023, 333: 120565.
[12] LIANG C, WANG W G, ZHOU T F, et al.Local-global context aware transformer for language-guided video segmentation[J]. IEEE transactions on pattern analysis and machine intelligence, 2023, 45(8): 10055-10069.
[13] 张东英, 代悦, 张旭, 等. 风电爬坡事件研究综述及展望[J]. 电网技术, 2018, 42(6): 1783-1792.
ZHANG D Y, DAI Y, ZHANG X, et al.Review and prospect of research on wind power ramp events[J]. Power system technology, 2018, 42(6): 1783-1792.
[14] HAN W Y, ZHU T, CHEN L M, et al.MCformer: multivariate time series forecasting with mixed-channels Transformer[J]. IEEE internet of things journal, 2024, 11(17): 28320-28329.
[15] CUTURI M, BLONDEL M.Soft-DTW: a differentiable loss function for time-series[J]. arXiv: 1703.01541.
[16] VINCENT L, THOME N.Shape and time distortion loss for training deep time series forecasting models[C]//Proceedings of the 33rd International Conference on Neural Information Processing Systems Vancouver. Canada, 2019: 4189-4201.
[17] XIE W Y, LI X-H, CAO C C, et al.ViT-CX: causal explanation of vision transformers[C]//Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence. Macau, China, 2023: 1569-1577.
[18] NIE Y Q, NGUYEN N H, SINTHONG P, et al.A time series is worth 64 words: long-term forecasting with transformers[J]. arXiv: 2211.14730.
[19] LU H, YE H J, ZHAN D C.The capacity and robustness trade-off: revisiting the channel independent strategy for multivariate time series forecasting[J]. IEEE transactions on knowledge and data engineering, 2024, 36(11): 7129-7142.
[20] LONG H, SANG L W, WU Z J, et al.Image-based abnormal data detection and cleaning algorithm via wind power curve[J]. IEEE transactions on sustainable energy, 2020, 11(2): 938-946.
[21] ZHANG B, ZHANG Y, LIU J, WANG B.FGFF descriptor and modified hu moment-based hand gesture recognition[J]. Sensors, 2021, 21(19): 6525.
[22] LEE H, LEE C,LIM H, et al.TILDE-Q: a transformation invariant loss function for time-series forecasting[J]. arXiv:2210.15050(2024).
[23] MO S T, WANG H X, LI B X, et al.TimeSQL: improving multivariate time series forecasting with multi-scale patching and smooth quadratic loss[J]. Information sciences, 2024, 671: 120652.
[24] CHEN Y B, XU J J.Solar and wind power data from the Chinese state grid renewable energy generation forecasting competition[J]. Scientific data, 2022, 9: 577.
[25] MENG Y, YUN S N, ZHAO Z N, et al.Short-term electricity load forecasting based on a novel data preprocessing system and data reconstruction strategy[J]. Journal of building engineering, 2023, 77: 107432.
[26] ZHAO Z N, YUN S N, JIA L Y, et al.Hybrid VMD-CNN-GRU-based model for short-term forecasting of wind power considering spatio-temporal features[J]. Engineering applications of artificial intelligence, 2023, 121: 105982.