为提高光伏发电功率预测的准确性,提出一种改进的Transformer概率预测方法,包括数据预处理、预测模型及后期处理流程。针对光伏发电功率数据缺失问题,提出预测均值引导的随机森林插补法对缺失数据进行插补,提升数据完整性。然后引入基于改进Transformer模型的概率预测方法,该模型利用多头注意力机制并结合归一化层和残差连接,可强化模型的鲁棒性和对长序列依赖问题的处理能力。在后处理阶段,结合四阶多项式和长短期记忆（LSTM）用于修正预测误差。最后使用历史数据进行实验验证,结果表明所提模型具有高预测准确性和可靠性。
Abstract
To improve the accuracy of photovoltaic power prediction, an improved Transformer probability prediction method is proposed, including data preprocessing, prediction model, and post-processing process. Given the problem of missing photovoltaic power data, this paper proposes a random forest interpolation method guided by the predicted mean to interpolate the missing data and improve the data integrity. Then, a probability prediction method based on the enhanced Transformer model is introduced. The model uses a multi-head attention mechanism combined with a normalization layer and a residual connection to strengthen the robustness of the model and its ability to handle long sequence dependency problems. In the post-processing stage, a fourth-order polynomial and LSTM are combined to correct the prediction error. Finally, historical data are used for experimental verification, and the results show that the proposed model has high prediction accuracy and reliability.
关键词
光伏发电 /
深度学习 /
建模 /
概率 /
插补 /
预测 /
长短期记忆
Key words
PV power /
deep learning /
modelling /
probability /
interpolation /
forecasting /
LSTM
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 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.
[2] PERSSON C, BACHER P, SHIGA T, et al.Multi-site solar power forecasting using gradient boosted regression trees[J]. Solar energy, 2017, 150: 423-436.
[3] LIU D, SUN K.Random forest solar power forecast based on classification optimization[J]. Energy, 2019, 187: 115940.
[4] SHI J, LEE W J, LIU Y Q, et al.Forecasting power output of photovoltaic systems based on weather classification and support vector machines[J]. IEEE transactions on industry applications, 2012, 48(3): 1064-1069.
[5] BABALHAVAEJI A, RADMANESH M, JALILI M, et al.Photovoltaic generation forecasting using convolutional and recurrent neural networks[J]. Energy reports, 2023, 9: 119-123.
[6] 秦宇, 许野, 王鑫鹏, 等. 基于改进FCM-LSTM的光伏出力短期预测研究[J]. 太阳能学报, 2024, 45(8): 304-313.
QIN Y, XU Y, WANG X P, et al.Study on short-term photovoltaic output prediction based on improved FCM-LSTM[J]. Acta energiae solaris sinica, 2024, 45(8): 304-313.
[7] ABDEL-NASSER M, MAHMOUD K.Accurate photovoltaic power forecasting models using deep LSTM-RNN[J]. Neural computing and applications, 2019, 31(7): 2727-2740.
[8] 陈庆明, 廖鸿飞, 孙颖楷, 等. 基于GWO-GRU的光伏发电功率预测[J]. 太阳能学报, 2024, 45(7): 438-444.
CHEN Q M, LIAO H F, SUN Y K, et al.Photovoltaic power prediction model based on GWO-GRU[J]. Acta energiae solaris sinica, 2024, 45(7): 438-444.
[9] 吴凡曈, 杨俊华, 杨梦丽, 等. 基于卷积门控循环单元的波浪发电系统输出功率预测[J]. 太阳能学报, 2024, 45(8): 682-688.
WU F T, YANG J H, YANG M L, et al.Output power prediction of wave power generation system based on convolutional gated cyclic unit[J]. Acta energiae solaris sinica, 2024, 45(8): 682-688.
[10] JURADO M, SAMPER M, ROSÉS R. An improved encoder-decoder-based CNN model for probabilistic short-term load and PV forecasting[J]. Electric power systems research, 2023, 217: 109153.
[11] QU J Q, QIAN Z, PEI Y.Day-ahead hourly photovoltaic power forecasting using attention-based CNN-LSTM neural network embedded with multiple relevant and target variables prediction pattern[J]. Energy, 2021, 232: 120996.
[12] LI Y Y, SONG L D, ZHANG S, et al.A TCN-based hybrid forecasting framework for hours-ahead utility-scale PV forecasting[J]. IEEE transactions on smart grid, 2023, 14(5): 4073-4085.
[13] 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.
[14] KIM T, KO W, KIM J.Analysis and impact evaluation of missing data imputation in day-ahead PV generation forecasting[J]. Applied sciences, 2019, 9(1): 204.
[15] KIM J, OBREGON J, PARK H, et al.Multi-step photovoltaic power forecasting using transformer and recurrent neural networks[J]. Renewable and sustainable energy reviews, 2024, 200: 114479.
[16] CHEN B, LI J H.Combined probabilistic forecasting method for photovoltaic power using an improved Markov chain[J]. IET generation, transmission & distribution, 2019, 13(19): 4364-4373.
[17] RAMAKRISHNA R, SCAGLIONE A, VITTAL V, et al.A model for joint probabilistic forecast of solar photovoltaic power and outdoor temperature[J]. IEEE transactions on signal processing, 2019, 67(24): 6368-6383.
[18] GNEITING T, LERCH S, SCHULZ B.Probabilistic solar forecasting: benchmarks, post-processing, verification[J]. Solar energy, 2023, 252: 72-80.
基金
新能源电力系统全国重点实验室2024年开放课题（LAPS24006）; 宁夏回族自治区自然科学基金（2023AAC03857）
PDF(2430 KB)
