针对光伏电站空间位置分散及用户之间数据缺失易导致的功率预测精度不足的问题,提出一种场景分类与隐私保护下的分布式光伏短期功率预测协同训练策略。首先,利用皮尔逊相关系数提取重要气象特征,并采用模糊C均值聚类（FCM）算法将历史数据集聚类划分为晴天、阴天和雨天。其次,将天气变化相似的区域聚类成若干组,判别、筛选同类别下晴天、非晴天集合,构建不同场景下光伏功率预测模型。然后,在一般的联邦学习迭代算法的基础上添加多任务学习算法,建立一种新型多任务模式的本地训练方法,保留参与联合建模的各光伏电站间的差异性。最后,对待测日进行预测,将其数据输入至上述建立的对应场景预测模型下,得到待测日的光伏功率预测结果。实验结果表明：在不同天气条件下,所提预测方法与多种网络模型相比,准确率最大可提升24.77%,均方根误差RMSE最大降低89.24%。与传统联邦框架相比,所提方法能在更快的训练轮次内达到目标用户识别率（UA）,缩短50%的通信轮数并且使平均UA提升8%。所提方案不仅在提高光伏短期功率预测的准确性方面得到验证,同时还展现出较强的适应性和鲁棒性。
Abstract
Due to the spatial dispersion of photovoltaic power stations and the lack of data sharing among users, insufficient power prediction accuracy commonly arises. This paper proposes a collaborative training strategy for distributed PV short-term power prediction based on scenario classification and privacy protection. Firstly, the Pearson correlation coefficient is used to extract important meteorological features, and the fuzzy C-means clustering （fuzzy C-means, FCM） algorithm is used to cluster the historical data set into sunny days, cloudy days and rainy days. Secondly, the regions with similar weather patterns are clustered into several groups, and the sets of sunny and non-sunny under the same category are discriminated and screened to build photovoltaic power prediction models under different scenarios. Then, a multi-task learning algorithm is added on the basis of the general iterative algorithm of federated learning, and a new local training method of multi-task mode is established to preserve the differences among the PV power stations participating in the joint modeling. Finally, the test day is predicted, and the data is input into the prediction model of the corresponding scenario established above, and the photovoltaic power prediction results of the test day are obtained. The experimental findings revealed that, under different weather conditions, compared with various network models, the accuracy of the proposed prediction method is increased by 24.77%, and the root-mean-square error （RMSE） is reduced by 89.24%. Compared with the traditional federated framework, the proposed scheme can achieve the target user identification rate （UA） in faster training rounds, shorten the number of communication rounds by 50% and increase the average UA by 8%. It is verified that this scheme not only improves the accuracy of photovoltaic short-term power prediction, but also has strong adaptability and robustness.
关键词
分布式光伏 /
功率预测 /
聚类分析 /
多任务联邦学习 /
场景分类
Key words
distributed photovoltaics /
power prediction /
cluster analysis /
multi-task federated learning /
scenario classification
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] LUO X, ZHANG D X, ZHU X.Combining transfer learning and constrained long short-term memory for power generation forecasting of newly-constructed photovoltaic plants[J]. Renewable energy, 2022, 185: 1062-1077.
[2] 万灿, 宋永华. 新能源电力系统概率预测理论与方法及其应用[J]. 电力系统自动化, 2021, 45(1): 2-16.
WAN C, SONG Y H.Theories, methodologies and applications of probabilistic forecasting for power systems with renewable energy sources[J]. Automation of electric power systems, 2021, 45(1): 2-16.
[3] LI P T, ZHOU K L, LU X H, et al.A hybrid deep learning model for short-term PV power forecasting[J]. Applied energy, 2020, 259: 114216.
[4] LIN G Q, LI L L, TSENG M L, et al.An improved moth-flame optimization algorithm for support vector machine prediction of photovoltaic power generation[J]. Journal of cleaner production, 2020, 253: 119966.
[5] 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.Research on short term prediction of photovoltaic output based on improved FCM-LSTM[J]. Acta energiae solaris sinica, 2024, 45(8): 304-313.
[6] 乔颖, 孙荣富. 基于数据增强的分布式光伏电站群短期功率预测(二): 网格化预测[J]. 电网技术, 2021, 45(6): 2210-2218.
QIAO Y, SUN R F.Distributed photovoltaic station cluster short-term power forecasting part II: gridding prediction[J]. Power system technology, 2021, 45(6): 2210-2218.
[7] ALAMIN Y I, ANATY M K, ÁLVAREZ HERVÁS J D, et al. Very short-term power forecasting of high concentrator photovoltaic power facility by implementing artificial neural network[J]. Energies, 2020, 13(13): 3493.
[8] AL-DAHIDI S, AYADI O, ADEEB J, et al.Extreme learning machines for solar photovoltaic power predictions[J]. Energies, 2018, 11(10): 2725.
[9] KUSHWAHA V, PINDORIYA N M.A SARIMA-RVFL hybrid model assisted by wavelet decomposition for very short-term solar PV power generation forecast[J]. Renewable energy, 2019, 140: 124-139.
[10] HUANG Q, WEI S Y.Improved quantile convolutional neural network with two-stage training for daily-ahead probabilistic forecasting of photovoltaic power[J]. Energy conversion and management, 2020, 220: 113085.
[11] 王开艳, 杜浩东, 贾嵘, 等. 基于相似日聚类和QR-CNN-BiLSTM模型的光伏功率短期区间概率预测[J]. 高电压技术, 2022, 48(11): 4372-4388.
WANG K Y, DU H D, JIA R, et al.Short-term interval probability prediction of photovoltaic power based on similar daily clustering and QR-CNN-BiLSTM model[J]. High voltage engineering, 2022, 48(11): 4372-4388.
[12] 管霖, 赵琦, 周保荣, 等. 基于多尺度聚类分析的光伏功率特性建模及预测应用[J]. 电力系统自动化, 2018, 42(15): 24-30.
GUAN L, ZHAO Q, ZHOU B R, et al.Multi-scale clustering analysis based modeling of photovoltaic power characteristics and its application in prediction[J]. Automation of electric power systems, 2018, 42(15): 24-30.
[13] BÖÖK H, LINDFORS A V. Site-specific adjustment of a NWP-based photovoltaic production forecast[J]. Solar energy, 2020, 211: 779-788.
[14] DENG F M, ZENG Z Q, MAO W, et al.A novel transmission line defect detection method based on adaptive federated learning[J]. IEEE transactions on instrumentation and measurement, 2023, 72: 1-12.
[15] 吴浩天, 孙荣富, 廖思阳, 等. 基于改进气象聚类分型的短期风电功率概率预测方法[J]. 电力系统自动化, 2022, 46(15): 56-65.
WU H T, SUN R F, LIAO S Y, et al.Short-term wind power probability forecasting method based on improved meteorological clustering and classification[J]. Automation of electric power systems, 2022, 46(15): 56-65.
[16] WANG H X, SHEN H T, LI F, et al.Novel PV power hybrid prediction model based on FL co-training method[J]. Electronics, 2023, 12(3): 730.
[17] ZHOU X, PANG C X, ZENG X H, et al.A short-term power prediction method based on temporal convolutional network in virtual power plant photovoltaic system[J]. IEEE transactions on instrumentation and measurement, 2023, 72: 1-10.
[18] TAN A Z, YU H, CUI L Z, et al.Towards personalized federated learning[J]. IEEE transactions on neural networks and learning systems, 2023, 34(12): 9587-9603.
[19] 张研, 景超, 王慧民, 等. 基于周期注意力机制的中长期光伏发电功率预测[J]. 太阳能学报, 2024, 45(10): 298-308.
ZHANG Y, JING C, WANG H M, et al.Medium and long term photovoltaic power prediction based on periodic attention mechanism[J]. Acta energiae solaris sinica, 2024, 45(10): 298-308.
[20] CHEN J B, LI J P, HUANG R Y, et al.Federated transfer learning for bearing fault diagnosis with discrepancy-based weighted federated averaging[J]. IEEE transactions on instrumentation measurement, 2022, 71: 3180417.
[21] LI H, REN Z Y, XU Y, et al.A multi-data driven hybrid learning method for weekly photovoltaic power scenario forecast[J]. IEEE transactions on sustainable energy, 2022, 13(1): 91-100.
[22] WANG F, XUAN Z M, ZHEN Z, et al.A day-ahead PV power forecasting method based on LSTM-RNN model and time correlation modification under partial daily pattern prediction framework[J]. Energy conversion and management, 2020, 212: 112766.
[23] 赵康宁, 蒲天骄, 王新迎, 等. 基于改进贝叶斯神经网络的光伏出力概率预测[J]. 电网技术, 2019, 43(12): 4377-4386.
ZHAO K N, PU T J, WANG X Y, et al.Probabilistic forecasting for photovoltaic power based on improved Bayesian neural network[J]. Power system technology, 2019, 43(12): 4377-4386.
[24] 冯裕祺,李辉,李利娟,等.基于CNN-GRU的光伏电站电压轨迹预测[J].中国电力,2022,55(7):163-171.
FENG Y Q, LI H, LI L J.et al.Voltage trajectory prediction of photovoltaic power station based on CNN-GRU[J]. Electric power, 2022, 55(7): 163-171.
[25] 赵晋川. 融合AdaBoost和GRU方法的地质分层识别研究[D]. 北京: 华北电力大学, 2019.
ZHAO J C.Research on geological stratification recognition based on AdaBoost and GRU method[D]. Beijing: North China Electric Power University, 2019.
[26] 时培明, 郭轩宇, 杜清灿, 等. 基于TCN-BiLSTM-Attention-ESN的光伏功率预测[J]. 太阳能学报, 2024, 45(9): 304-316.
SHI P M, GUO X Y, DU Q C, et al.Photovoltaic power prediction based on TCN-BiLSTM-Attention-ESN[J]. Acta energiae solaris sinica, 2024, 45(9): 304-316.
[27] YU F, LIN H, WANG X D, et al.Communication-efficient personalized federated meta-learning in edge networks[J]. IEEE transactions on network and service management, 2023, 20(2): 1558-1571.
[28] MILLS J, HU J, MIN G Y.Multi-task federated learning for personalised deep neural networks in edge computing[J]. IEEE transactions on parallel and distributed systems, 2022, 33(3): 630-641.
[29] LI X J, WANG L J, YAN N, et al.Cooperative dispatch of distributed energy storage in distribution network with PV generation systems[J]. IEEE transactions on applied superconductivity, 2021, 31(8): 3117750.
基金
国家自然科学基金（52167008; 52377103）
PDF(2131 KB)
