基于贝叶斯优化超参数的光伏功率区间预测方法

陆佳恩; 赵健; 李小勇

doi:10.19912/j.0254-0096.tynxb.2024-1938

PDF(3196 KB)

太阳能学报 ›› 2026, Vol. 47 ›› Issue (3) : 644-655. DOI: 10.19912/j.0254-0096.tynxb.2024-1938

基于贝叶斯优化超参数的光伏功率区间预测方法

陆佳恩¹, 赵健¹, 李小勇²

作者信息 +

PHOTOVOLTAIC POWER INTERVAL PREDICTION METHOD BASED ON BAYESIAN OPTIMIZATION TO IMPROVE DEEP LEARNING MODEL HYPERPARAMETERS

Lu Jiaen¹, Zhao Jian¹, Li Xiaoyong²

Author information +

文章历史 +

摘要

针对光伏发电中光伏功率存在的随机、间歇和波动问题对电网安全的挑战,提出一种创新的深度学习模型——基于融合空时特征提取的深度残差时序卷积-高斯过程回归DRSTCG-GPR模型。该模型融合空时特征,在残差模块引入软阈值与注意力机制,通过贝叶斯算法自动优化超参数,并利用高斯过程回归量化预测的不确定性,最终实现短期光伏功率的高精度区间预测。实验对比表明,相较于传统卷积神经网络（CNN）、门控循环单元（GRU）及卷积门控循环单元（CGRU）模型,该模型在点预测精度（R²提升2.20%）、区间预测覆盖性能（I_CP提升2.10%）及概率预测可靠性上均展现出显著优势。

Abstract

To address the challenges to grid security posed by the randomness, intermittency, and fluctuations of photovoltaic （PV） power in PV power generation planning, an innovative deep learning model—the DRSTCG-GPR model based on fused spatiotemporal feature extraction—is proposed. This model fuses spatiotemporal features, introduces soft thresholding and attention mechanisms into the residual module, automatically optimizes hyperparameters via a Bayesian algorithm, and quantifies the uncertainty of predictions using Gaussian process regression, ultimately achieving high-precision interval prediction for short-term photovoltaic （PV） power.. Experimental comparisons show that, compared to traditional CNN, GRU, and CGRU models, this model exhibits significant advantages in point prediction accuracy （R² improvement of 2.20%）, interval prediction coverage performance （I_CP improvement of 2.10%）, and probabilistic prediction reliability.

导出引用

陆佳恩, 赵健, 李小勇. 基于贝叶斯优化超参数的光伏功率区间预测方法[J]. 太阳能学报. 2026, 47(3): 644-655 https://doi.org/10.19912/j.0254-0096.tynxb.2024-1938

Lu Jiaen, Zhao Jian, Li Xiaoyong. PHOTOVOLTAIC POWER INTERVAL PREDICTION METHOD BASED ON BAYESIAN OPTIMIZATION TO IMPROVE DEEP LEARNING MODEL HYPERPARAMETERS[J]. Acta Energiae Solaris Sinica. 2026, 47(3): 644-655 https://doi.org/10.19912/j.0254-0096.tynxb.2024-1938

中图分类号： TM615

参考文献

[1] 赖昌伟, 黎静华, 陈博, 等. 光伏发电出力预测技术研究综述[J]. 电工技术学报, 2019, 34(6): 1201-1217.
LAI C W, LI J H, CHEN B, et al.Review of photovoltaic power output prediction technology[J]. Transactions of China Electrotechnical Society, 2019, 34(6): 1201-1217.
[2] 李芬, 于淏, 孙改平, 等. 基于优化变分模态分解的光伏功率超短期区间预测方法[J]. 太阳能学报, 2024, 45(8): 367-376.
LI F, YU H, SUN G P, et al.Ultra short term interval prediction method of photovoltaic power based on optimal variational mode decomposition[J]. Acta energiae solaris sinica, 2024, 45(8): 367-376.
[3] 董存, 王铮, 白捷予, 等. 光伏发电功率超短期预测方法综述[J]. 高电压技术, 2023, 49(7): 2938-2951.
DONG C, WANG Z, BAI J Y, et al.Review of ultra-short-term forecasting methods for photovoltaic power generation[J]. High voltage engineering, 2023, 49(7): 2938-2951.
[4] 崔文康. 人工智能驱动的新能源发电功率非参数概率预测研究[D]. 杭州: 浙江大学, 2023.
CUI W K.Research on artificial intelligence-driven nonparametric probabilistic forecasting of renewable energy generation[D]. Hangzhou: Zhejiang University, 2023.
[5] LI F, LI C Y, YANG Y, et al.Short-term photovoltaic power probability forecasting based on OLPP-GPR and modified clearness index[J]. The journal of engineering, 2017, 2017(13): 1625-1628.
[6] WOLFF B, LORENZ E, KRAMER O.Statistical learning for short-term photovoltaic power predictions[M]//Computational Sustainability. Cham: Springer International Publishing, 2016: 31-45.
[7] MASSAOUDI M, CHIHI I, SIDHOM L, et al.Enhanced random forest model for robust short-term photovoltaic power forecasting using weather measurements[J]. Energies, 2021, 14(13): 3992.
[8] WOLFF B, KÜHNERT J, LORENZ E, et al. Comparing support vector regression for PV power forecasting to a physical modeling approach using measurement, numerical weather prediction, and cloud motion data[J]. Solar energy, 2016, 135: 197-208.
[9] MELLIT A, PAVAN A M, LUGHI V.Deep learning neural networks for short-term photovoltaic power forecasting[J]. Renewable energy, 2021, 172: 276-288.
[10] WU Y, GONG Y Z, CHEN X G, et al.Wind speed interval prediction based on Bayesian optimized spatio-temporal integration and compression deep residual network[J]. Sensors, 2025, 25(20): 6370.
[11] FAN J, ZHANG K, HUANG Y P, et al.Parallel spatio-temporal attention-based TCN for multivariate time series prediction[J]. Neural computing and applications, 2023, 35(18): 13109-13118.
[12] 张程, 林谷青, 匡宇. 基于MEEMD-QUATRE-BILSTM的短期光伏出力区间预测[J]. 太阳能学报, 2023, 44(11): 40-54.
ZHANG C, LIN G Q, KUANG Y.Short-term PV output interval prediction based on MEEMD-QUATRE-BILSTM[J]. Acta energiae solaris sinica, 2023, 44(11): 40-54.
[13] 宋绍剑, 罗世坚, 李国进, 等. 基于MPA-LSTM模型和Bootstrap方法的短期光伏功率区间预测[J]. 广西大学学报(自然科学版), 2022, 47(4): 986-997.
SONG S J, LUO S J, LI G J, et al.A short-term PV power interval forecasting based on MPA-LSTM network model and Bootstrap method[J]. Journal of Guangxi University (natural science edition), 2022, 47(4): 986-997.
[14] ANTONANZAS J, OSORIO N, ESCOBAR R, et al.Review of photovoltaic power forecasting[J]. Solar energy, 2016, 136: 78-111.
[15] NI Q, ZHUANG S X, SHENG H M, et al.An ensemble prediction intervals approach for short-term PV power forecasting[J]. Solar energy, 2017, 155: 1072-1083.
[16] LIMOUNI T, YAAGOUBI R, BOUZIANE K, et al.Accurate one step and multistep forecasting of very short-term PV power using LSTM-TCN model[J]. Renewable energy, 2023, 205: 1010-1024.
[17] EHTERAM M, AFSHARI NIA M, PANAHI F, et al.Gaussian mutation-orca predation algorithm-deep residual shrinkage network(DRSN)-temporal convolutional network (TCN)-random forest model: an advanced machine learning model for predicting monthly rainfall and filtering irrelevant data[J]. Environmental sciences Europe, 2024, 36(1): 13.
[18] GUANOLUISA R, ARCOS-AVILES D, FLORES-CALERO M, et al.Photovoltaic power forecast using deep learning techniques with hyperparameters based on Bayesian optimization: a case study in the Galapagos Islands[J]. Sustainability, 2023, 15(16): 12151.
[19] WAN X, LI X Y, WANG X Z, et al.Water quality prediction model using Gaussian process regression based on deep learning for carbon neutrality in papermaking wastewater treatment system[J]. Environmental research, 2022, 211: 112942.
[20] HE J A, LI X Y, CHEN Z G, et al.A hybrid CLSTM-GPR model for forecasting particulate matter(PM_2.5)[J]. Atmospheric pollution research, 2023, 14(8): 101832.
[21] 郑洪庆, 宋万清, 江月松, 等. 长相关随机模型FBM对光伏发电短期预测[J]. 电网与清洁能源, 2024, 40(4): 102-111.
ZHENG H Q, SONG W Q, JIANG Y S, et al.Short-term forecasting of photovoltaic power generation by the long correlation stochastic model FBM[J]. Power system and clean energy, 2024, 40(4): 102-111.
[22] 伍耘, 葛佳敏, 王文烨, 等. 贝叶斯优化超参数的空时融合压缩残差网络在风速区间预测中的研究[J]. 电力系统保护与控制, 2025, 53(1): 13-23.
WU Y, GE J M, WANG W Y, et al.Wind speed interval prediction using spatio-temporal fusion compressed residual networks with Bayesian optimized hyperparameters[J]. Power system protection and control, 2025, 53(1): 13-23.
[23] LI X Y, YI X H, LIU Z H, et al.Application of novel hybrid deep leaning model for cleaner production in a paper industrial wastewater treatment system[J]. Journal of cleaner production, 2021, 294: 126343.
[24] IN X B, ZHENG W Z, KONG J L, et al.Deep-learning forecasting method for electric power load via attention-based encoder-decoder with Bayesian optimization[J]. Energies, 2021, 14(6): 1596.