对比分析单一模型、堆叠型和并联型混合深度学习网络在风电功率超短期预测中的表现,构建基于长短期记忆网络（LSTM）、时间卷积网络（TCN）与循环神经网络（RNN）的混合预测模型（包括堆叠型架构模型、混合并联型架构模型和双并联型架构模型）。实验结果表明,并联型架构深度学习网络风电功率预测模型在预测精度和计算效率方面显著优于堆叠型架构模型。在相同超参数配置下,并联架构的PA-TCN-LSTM模型在9种模型中表现最佳,相比对应堆叠型架构模型,平均绝对误差（MAE）降低10.89%,均方根误差（RMSE）降低5.96%,决定系数（R2）提升0.35%,训练时间缩短31.99%。相较于单一模型,PA-TCN-LSTM模型的MAE平均降低27.76%,RMSE平均降低35.17%,R2平均提升1.20%。进一步,基于TPE贝叶斯优化对混合并联型架构与双并联型架构模型的预测性能和效率进行对比分析,结果表明在实验构建的对比模型中,TPE-PA-TCN-LSTM在预测精度和训练效率方面表现最佳。
Abstract
This paper compares and analyzes the performance of single models, stacked hybrid deep learning networks, and parallel hybrid deep learning networks for ultrashortterm wind power forecasting. Hybrid forecasting models based on Long ShortTerm Memory （LSTM）, Temporal Convolutional Networks （TCN）, and Recurrent Neural Networks （RNN） are developed, including stacked architectures, hybrid parallel architectures, and dualparallel architectures. Experimental results show that parallelarchitecture deep learning networks for wind power forecasting significantly outperform stackedarchitecture models in terms of prediction accuracy and computational efficiency. Under the same hyperparameter configuration, the parallelarchitecture PATCNLSTM model achieves the best performance among nine models. Compared with the corresponding stackedarchitecture model, it reduces the mean absolute error （MAE） by 10.89%, the root mean square error （RMSE） by 5.96%, increases the coefficient of determination （R2） by 0.35%, and shortens the training time by 31.99%. Compared with single models, the PATCNLSTM model reduces MAE by 27.76% on average, RMSE by 35.17% on average, and increases R2 by 1.20% on average. Furthermore, a comparative analysis of the prediction performance and efficiency of hybrid parallel and dualparallel architecture models is conducted using TPE Bayesian optimization. The results indicate that among the compared models developed in this study, TPEPATCNLSTM achieves the best prediction accuracy and training efficiency.
关键词
风电功率 /
预测模型 /
深度学习 /
堆叠深度学习模型 /
并联深度学习模型 /
超参数优化
Key words
wind power /
prediction models /
deep learning /
stacked deep learning model /
parallel deep learning model /
hyperparameter optimization
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] Faruque M O, Hossain M A, Islam M R, et al.Very short-term wind power forecasting for real-time operation using hybrid deep learning model with optimization algorithm[J]. Cleaner Energy Systems, 2024, 9: 100129.
[2] 付文龙, 章轩瑞, 张海荣, 等. 基于INGO-SWGMN混合模型的超短期风速预测研究[J]. 太阳能学报, 2024, 45(5): 133-143.
Fu W L, Zhang X R, Zhang H R, et al.Ultra-short-term wind speed prediction based on INGO-SWGMN hybrid model[J]. Acta Energiae Solaris Sinica, 2024, 45(5): 133-143.
[3] 国家能源局. 国家能源局发布2025年1-2月份全国电力工业统计数据[EB/OL]. (2025-03-20)[2026-05-06].
National Energy Administration. National Energy Administration releases national power industry statistics for January-February2025[EB/OL]. (2025-03-20)[2026-05-06].
[4] 李沛智. 基于深度学习的短期风电功率预测[D]. 太原: 太原科技大学, 2024.
Li P Z.Short term wind power forecasting based on deep learning[D]. Taiyuan: Taiyuan University of Science and Technology, 2024.
[5] 刘谭, 刘娜, 刘贵平, 等. 深度学习方法在风电功率预测中的应用与研究方向概述[J]. 计算机科学与探索, 2025, 19(3): 602-622.
Liu T, Liu N, Liu G P, et al.Overview of applications and research directions of deep learning methods for wind power prediction[J]. Journal of Frontiers of Computer Science & Technology, 2025, 19(3): 602-622.
[6] 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.
[7] Wang Y, Zou R M, Liu F, et al.A review of wind speed and wind power forecasting with deep neural networks[J]. Applied Energy, 2021, 304: 117766.
[8] 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.
[9] Carolin Mabel M, Fernandez E.Analysis of wind power generation and prediction using ANN: a case study[J]. Renewable Energy, 2008, 33(5): 986-992.
[10] 杨秀媛, 裘微江, 金鑫城, 等. 改进K近邻算法在风功率预测及风水协同运行中的应用[J]. 电网技术, 2018, 42(3): 772-778.
Yang X Y, Qiu W J, Jin X C, et al.Wind power prediction based on improved K-nearest neighbor algorithm and its application in co-operation of wind and hydro powers[J]. Power System Technology, 2018, 42(3): 772-778.
[11] Dhiman H S, Deb D, Guerrero J M.Hybrid machine intelligent SVR variants for wind forecasting and ramp events[J]. Renewable and Sustainable Energy Reviews, 2019, 108: 369-379.
[12] 罗潇远, 刘杰, 杨斌, 等. 基于改进鱼鹰优化算法与VMD-LSTM的超短期风电功率预测[J]. 太阳能学报, 2025, 46(3): 652-660.
Luo X Y, Liu J, Yang B, et al.Ultra short-term wind power prediction based on improved ospery optimization algorithm and VMD-LSTM[J]. Acta Energiae Solaris Sinica, 2025, 46(3): 652-660.
[13] Kisvari A, Lin Z, Liu X L.Wind power forecasting-a data-driven method along with gated recurrent neural network[J]. Renewable Energy, 2021, 163: 1895-1909.
[14] Lara-Benítez P, Carranza-García M, Luna-Romera J M, et al. Temporal convolutional networks applied to energy-related time series forecasting[J]. Applied Sciences, 2020, 10(7): 2322.
[15] 王铮, Rui Pestana, 冯双磊, 等. 基于加权系数动态修正的短期风电功率组合预测方法[J]. 电网技术, 2017, 41(2): 500-507.
Wang Z, Pestana R, Feng S L, et al.Short-term wind power combination forecasting method based on dynamic coefficient updating[J]. Power System Technology, 2017, 41(2): 500-507.
[16] 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.
[17] Zhang X D, Yang M J, Liu N, et al.Wind power error compensation prediction model based on CEEMD-SE-ELM-TCN[J]. International Journal of Low-Carbon Technologies, 2024, 19: 972-979.
[18] Wang J J, Liu Y F, Li Y N.A parallel differential learning ensemble framework based on enhanced feature extraction and anti-information leakage mechanism for ultra-short-term wind speed forecast[J]. Applied Energy, 2024, 361: 122909.
[19] Huynh D N.Wind power curve modeling[EB/OL]. https://www.kaggle.com/code/winternguyen/wind-power-curve-modeling/input.
[20] Schafer R W.What is a savitzky-golay filter? [lecture notes[J]. IEEE Signal Processing Magazine, 2011, 28(4): 111-117.
[21] Liu S J, Zhang Y C, Du X Z, et al.Short-term power prediction of wind turbine applying machine learning and digital filter[J]. Applied Sciences, 2023, 13(3): 1751.
[22] Zuo B, Cheng J S, Zhang Z H.Degradation prediction model for proton exchange membrane fuel cells based on long short-term memory neural network and Savitzky-Golay filter[J]. International Journal of Hydrogen Energy, 2021, 46(29): 15928-15937.
[23] 李庆春, 潘作枢, 李九亮. 变阶数滑动趋势分析及其应用[J]. 石油物探, 1994, 33(3): 92-98, 114.
Li Q C, Pan Z S, Li J L.Order-variant sliding trend analysis and its application[J]. Geophysical Prospecting for Petrole, 1994, 33(3): 92-98, 114.
[24] Newcomer M W, Hunt R J.NWTOPT-A hyperparameter optimization approach for selection of environmental model solver settings[J]. Environmental Modelling & Software, 2022, 147: 105250.
[25] 查雯婷, 闫利成, 陈波, 等. 基于TPE-LSTM的区域超短期风电功率预测[J]. 计算机应用与软件, 2022, 39(11): 25-30, 111.
Zha W T, Yan L C, Chen B, et al.Regional ultra-short-term wind power forecasting method based on TPE-LSTM[J]. Computer Applications and Software, 2022, 39(11): 25-30, 111.
基金
国家自然科学基金（52366009; 52130607）; 甘肃重点研发计划（26YFGA029）; 兰州市青年科技人才创新项目（2025-QN-078）
PDF(1453 KB)
