针对风电场数据具有复杂的非线性关系,且难以捕捉长距离依赖关系的问题,引入时间段分割策略,提出一种新的基于贝叶斯参数优化Transformer时间段分割模型的超短期风电功率预测方法（PBY-Trans）。该方法采用时间段分割技术分割风电场数据为子序列,将其作为Transformer模型编码器的输入,从而更好地适应时间序列的非线性特征;使用贝叶斯算法搜索Transformer模型参数的最优配置,实现模型性能提升,进一步提高预测准确度。采用Bengaluru某风电场数据集对所提方法的预测性能进行比较验证,相较于SVM、RNN、Informer、LSTM、GRU和TCN模型,所提PBY-Trans方法在平均绝对误差（MAE）分别下降33.56%、59.75%、47.27%、32.34%、40.46%和27.71%,均方根误差（RMSE）分别下降68.99%、37.05%、27.60%、14.43%、16.42%和12.29%。结果表明,该文提出的Transform时间段分割模型的风电功率预测方法能够进一步提高预测精度,鲁棒性能得到有效提升。
Abstract
Aiming at the complex nonlinear relationship and the difficulty in capturing long-range dependencies of wind farm data, a new ultra-short-term wind power prediction method （PBY-Trans） based on a Bayesian parameter-optimized Transformer time segmenting model is proposed by introducing a time segmentation strategy. This method employs the time segmentation technique to divide the wind farm data into subsequences, which are then utilized as the inputs to the Transformer model encoder to better adapt to the nonlinear characteristics of the time series. Furthermore, a Bayesian algorithm is employed to search for the optimal configuration of the Transformer model parameters, thereby enhancing the model performance and improving the prediction accuracy. The prediction performance of the proposed method is compared and verified using the data set of a wind farm in Bengaluru. Compared with the SVM, RNN, Informer, LSTM, GRU and TCN models, the mean absolute error（MAE） metric of the proposed PBY-Trans method achieved a decrease of 33.56%, 59.75%, 47.27%, 32.34%, 40.46% and 27.71%, and the root mean square error（RMSE） was also reduced by 68.99%, 37.05%, 27.60%, 14.43%, 16.42% and 12.29%, respectively. The results indicate that the proposed PBY-Trans prediction model can further enhance prediction accuracy and the robustness can be effectively enhanced.
关键词
风电功率 /
预测模型 /
贝叶斯优化 /
Transformer模型 /
时间段分割
Key words
wind power /
prediction model /
Bayesian optimization /
Transformer model /
time segmentation
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 钟雅珊, 付聪, 钱峰, 等. 考虑广义储能和条件风险价值的综合能源系统经济调度[J]. 电力系统保护与控制, 2022, 50(9): 54-63.ZHONG Y S, FU C, QIAN F, et al. Economic dispatch model of an integrated energy system considering generalized energy storage and conditional value at risk[J]. Power system protection and control, 2022, 50(9): 54-63.
[2] 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 H P, LI H, KAN T Y, et al. DWT-DTCNA ultra-short-term wind power prediction considering wind power timing characteristics[J]. Power system technology, 2023, 47(4): 1653-1665.
[3] JUNG J, BROADWATER R P.Current status and future advances for wind speed and power forecasting[J]. Renewable and sustainable energy reviews, 2014, 31: 762-777.
[4] 向阳, 刘亚娟, 孙志伟, 等. 基于帝王蝶算法的CNN-GRU-LightGBM模型短期风电功率预测[J]. 太阳能学报, 2025, 46(1): 105-114.XIANG Y, LIU Y J, SUN Z W, et al. Short term wind power prediction using CNN-GRU-LightGBM model based on emperor butterfly algorithm[J]. Acta energiae solaris sinica, 2025, 46(1): 105-114.
[5] HAN Q K, MENG F M, HU T, et al.Non-parametric hybrid models for wind speed forecasting[J]. Energy conversion and management, 2017, 148: 554-568.
[6] CORTES C, VAPNIK V.Support-vector networks[J]. Machine learning, 1995, 20(3): 273-297.
[7] 刘擘龙, 张宏立, 王聪, 等. 基于序列到序列和注意力机制的超短期风速预测[J]. 太阳能学报, 2021, 42(9): 286-294.LIU B L, ZHANG H L, WANG C, et al. Ultra-short-term wind speed prediction based on sequence-to-sequence and attention mechanism[J]. Acta energiae solaris sinica, 2021, 42(9): 286-294.
[8] 王颉, 刘兴杰, 梁英, 等. 一种基于MGWO-Informer的超短期风电功率预测方法[J]. 太阳能学报, 2024, 45(11): 477-485.WANG J, LIU X J, LIANG Y, et al. An ultra-short-term wind power prediction method based on MGWO-Informer[J]. Acta energiae solaris sinica, 2024, 45(11): 477-485.
[9] LIPTON Z C, BERKOWITZ J, ELKAN C. A critical review of recurrent neural networks for sequence learning[EB/OL]. 2015: arXiv preprint arXiv: 1506.00019. https://arxiv.org/abs/1506.00019.
[10] LIU H, MI X W, LI Y F.Smart multi-step deep learning model for wind speed forecasting based on variational mode decomposition, singular spectrum analysis, LSTM network and ELM[J]. Energy conversion and management, 2018, 159: 54-64.
[11] NIU Z W, YU Z Y, TANG W H, et al.Wind power forecasting using attention-based gated recurrent unit network[J]. Energy, 2020, 196: 117081.
[12] HU H L, WANG L, LYU S X.Forecasting energy consumption and wind power generation using deep echo state network[J]. Renewable energy, 2020, 154: 598-613.
[13] GOODFELLOW I, POUGET-ABADIE J, MIRZA M, et al.Generative adversarial nets[C]//Advances in Neural Information Processing Systems. Montreal, Canada, 2014, 27: 2672-2680.
[14] BAI S J, KOLTER J Z, KOLTUN V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling[EB/OL]. 2018: arXiv: 1803.01271. https://arxiv.org/abs/1803.01271.
[15] LIM B, ARIK S Ö, LOEFF N, et al.Temporal Fusion Transformers for interpretable multi-horizon time series forecasting[J]. International journal of forecasting, 2021, 37(4): 1748-1764.
[16] VASWANI A, SHAZEER N, PARMAR N, et al.Attention is all you need[C]//Advances in Neural Information Processing Systems. Longbeach, America, 2017, 30: 5998-6008.
[17] KALYAN K S, RAJASEKHARAN A, SANGEETHA S. AMMUS: a survey of transformer-based pretrained models in natural language processing[EB/OL]. 2021: arXiv: 2108.05542. https://arxiv.org/abs/2108.05542.
[18] KHAN S, NASEER M, HAYAT M, et al.Transformers in vision: a survey[J]. ACM computing surveys, 2022, 54(10s): 1-41.
[19] KARITA S, CHEN N X, HAYASHI T, et al.A comparative study on transformer vs RNN in speech applications[C]//2019 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU). SG, Singapore, 2019: 449-456.
[20] WEN Q S, ZHOU T, ZHANG C L, et al. Transformers in time series: a survey[EB/OL]. 2022: arXiv: 2202.07125. https://arxiv.org/abs/2202.07125.
[21] ZHOU H Y, ZHANG S H, PENG J Q, et al.Informer: beyond efficient transformer for long sequence time-series forecasting[J]. Proceedings of the AAAI conference on artificial intelligence, 2021, 35(12): 11106-11115.
[22] WU H X, XU J H, WANG J M, et al.Autoformer: decomposition transformers with auto-correlation for long-term series forecasting[C]//Advances in Neural Information Processing Systems, 2021, 34: 22419-22430.
[23] ZHOU T, MA Z Q, WEN Q S, et al. FEDformer: frequency enhanced decomposed transformer for long-term series forecasting[EB/OL]. 2022: arXiv preprint arXiv: 2201.12740. https://arxiv.org/abs/2201.12740.
[24] ZENG A L, CHEN M X, ZHANG L, et al.Are transformers effective for time series forecasting?[J]. Proceedings of the AAAI conference on artificial intelligence, 2023, 37(9): 11121-11128.
[25] BASHIR T, WANG H F, TAHIR M, et al.Wind and solar power forecasting based on hybrid CNN-ABiLSTM, CNN-transformer-MLP models[J]. Renewable energy, 2025, 239: 122055.
[26] WANG D, XU M, ZHU G M, et al.Enhancing wind power forecasting accuracy through LSTM with adaptive wind speed calibration (C-LSTM)[J]. Scientific reports, 2025, 15: 5352.
[27] YU Y X, HAN X S, YANG M, et al.Probabilistic prediction of regional wind power based on spatiotemporal quantile regression[C]//2019 IEEE Industry Applications Society Annual Meeting. Baltimore, MD, USA, 2019: 1-16.
[28] ZHAO H S, ZHANG Y, LIU S, et al.PSANet: point-wise spatial attention network for scene parsing[C]//Computer Vision-ECCV 2018. Munich, Germany, 2018: 270-286.
[29] NIE Y Q, NGUYEN N H, SINTHONG P, et al. A time series is worth 64 words: long-term forecasting with transformers[EB/OL]. 2022: arXiv: 2211.14730. https://arxiv.org/abs/2211.14730.
[30] SNOEK J, LAROCHELLE H, ADAMS R P.Practical Bayesian optimization of machine learning algorithms[C]//Neural Information Processing Systems. Nevada, USA, 2012
[31] 王健峰, 张磊, 陈国兴, 等. 基于改进的网格搜索法的SVM参数优化[J]. 应用科技, 2012, 39(3): 28-31.WANG J F, ZHANG L, CHEN G X, et al. A parameter optimization method for an SVM based on improved grid search algorithm[J]. Applied science and technology, 2012, 39(3): 28-31.
[32] BERGSTRA J, BENGIO Y.Random search for hyper-parameter optimization[J]. The journal of machine learning research, 2012, 13: 281-305.
[33] 陈峰, 余轶, 徐敬友, 等. 基于Bayes-LSTM网络的风电出力预测方法[J]. 电力系统保护与控制, 2023, 51(6): 170-178.CHEN F, YU Y, XU J Y, et al. Prediction method of wind power output based on a Bayes-LSTM network[J]. Power system protection and control, 2023, 51(6): 170-178.
[34] 刘新宇, 蒲欣雨, 李继方, 等. 基于贝叶斯优化的VMD-GRU短期风电功率预测[J]. 电力系统保护与控制, 2023, 51(21): 158-165.LIU X Y, PU X Y, LI J F, et al. Short-term wind power prediction of a VMD-GRU based on Bayesian optimization[J]. Power system protection and control, 2023, 51(21): 158-165.
[35] 季廷炜, 莫邵昌, 谢芳芳, 等. 基于高斯过程回归的机翼/短舱一体化气动优化[J]. 浙江大学学报(工学版), 2023, 57(3): 632-642.JI T W, MO S C, XIE F F, et al. Integrated aerodynamic optimization of wing/nacelle based on Gaussian process regression[J]. Journal of Zhejiang University (engineering science), 2023, 57(3): 632-642.
基金
国家自然科学基金（51607004）; 安徽省自然科学基金（2008085MF197）
PDF(19874 KB)
