基于深度学习的集中式光伏电站在线增量功率预测方法

doi:10.11930/j.issn.1004-9649.202510014

摘要/Abstract

摘要：

在“双碳”目标下，集中式光伏电站已成为新能源电力系统的重要支撑，但光伏发电功率受季节、天气等因素影响具有强间歇性与波动性。针对实际应用中输入数据动态演变导致模型性能衰减，且传统更新方式易引发灾难性遗忘的问题，提出一种基于深度学习的在线增量功率预测模型。该模型引入深度经验回放++（deep experience replay，DER++）增量学习机制，构建“分块特征提取-在线知识保留”双核心框架，通过补丁令牌策略捕捉多尺度周期性特征，利用自注意力机制挖掘多变量依赖关系，结合经验回放技术缓解灾难性遗忘。基于某光伏电站实测数据表明，所提模型的累计精度衰减率远低于传统模型，展现出更强的适应性与泛化能力，为集中式光伏功率在线动态预测提供了有效解决方案。

关键词: 集中式光伏电站, 功率预测, 深度学习

Abstract:

Under the "dual carbon" goals, centralized photovoltaic (PV) power stations have become a crucial support for the new energy power syst however, PV power generation is strongly intermittent and volatile due to factors such as seasons and weather. Addressing the issue of model performance degradation caused by the dynamic evolution of input data in practical acations, and the susceptibility of traditional updating methods to catastrophic forgetting, this paper proposes a deep learning-based online incremental power prediction model. The model introduces the Deep Experience Replay (DER) incremenng mechanism to construct a dual-core framework of "block feature extraction and online knowledge retention". It captures multi-scale periodic features via a patch token strategy, utilizes self-attention mechanto mine multivariate dependencies, and combines experience replay techniques to alleviate catastrophic forgetting. Experimental results based on real-world data from a PV power station indicate that the cumulative accuracy degradation rate of the prsed model is significantly lower than that of traditional models, demonstrating stronger adaptability and generalization capabilities, and providing an effective solution for online dynamic power prediction in centralized.

Key words: centralized photovoltaic power station, power prediction, deep learning

杨超颖, 李慧蓬, 赵军. 基于深度学习的集中式光伏电站在线增量功率预测方法[J]. 中国电力, 2026, 59(5): 142-149.

YANG Chaoying, LI Huipeng, ZHAO Jun. Online incremental power forecasting method for centralized photovoltaic power plants based on deep learning[J]. Electric Power, 2026, 59(5): 142-149.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://www.electricpower.com.cn/CN/10.11930/j.issn.1004-9649.202510014

https://www.electricpower.com.cn/CN/Y2026/V59/I5/142

图/表 5

参考文献 35

1	王帅, 贾东梨, 刘科研, 等. 适应高比例光伏接入的有源配电网改进型故障保护策略[J]. 中国电力, 2026, 59 (1): 105- 114.
	WANG Shuai, JIA Dongli, LIU Keyan, et al. Improved fault protection strategy for active distribution networks adapting to high proportion photovoltaic access[J]. Electric Power, 2026, 59 (1): 105- 114.
2	贾东梨, 任昭颖, 刘科研, 等. 计及多种分布式能源的多端直流配电网故障电流计算方法[J]. 中国电力, 2026, 59 (1): 84- 96.
	JIA Dongli, REN Zhaoying, LIU Keyan, et al. Fault current calculation method for multi-terminal DC distribution networks considering multiple distributed generation[J]. Electric Power, 2026, 59 (1): 84- 96.
3	孙怡文, 邢海军, 梅丘梅, 等. 计及风电-光伏出力相关性的新型电力系统可靠性评估[J]. 浙江电力, 2025, 44 (9): 13- 20.
	SUN Yiwen, XING Haijun, MEI Qiumei, et al. Reliability assessment of modern power systems accounting for the correlation of wind-PV output correlation[J]. Zhejiang Electric Power, 2025, 44 (9): 13- 20.
4	王芊瑞, 阮景昕, 王跃社. 考虑风、光出力时空相关性的电-氢协同储能系统经济性优化调度研究[J]. 综合智慧能源, 2025, 47 (12): 34- 45.
	WANG Qianrui, RUAN Jingxin, WANG Yueshe. Economic optimal scheduling of electricity-hydrogen coordinated energy storage system considering spatiotemporal correlation of wind and photovoltaic power outputs[J]. Integrated Intelligent Energy, 2025, 47 (12): 34- 45.
5	郎益涛, 黄美珑, 吴沂林, 等. 光伏配储系统多场景优化调度策略研究[J]. 湖南电力, 2025, 45 (2): 30- 35.
	LANG Yitao, HUANG Meilong, WU Yilin, et al. Research on multi-scenario optimal scheduling strategy of photovoltaic distribution and storage system[J]. Hunan Electric Power, 2025, 45 (2): 30- 35.
6	李旭涛, 周洋, 常启诚, 等. 考虑电制氢辅助的配电网两阶段电压优化控制方法[J]. 中国电力, 2026, 59 (2): 81- 89.
	LI Xutao, ZHOU Yang, CHANG Qicheng, et al. Two-stage voltage optimization control method for hydrogen-production-assisted distribution networks[J]. Electric Power, 2026, 59 (2): 81- 89.
7	余笑东, 凌煦, 王源, 等. 基于改进全纯嵌入法的含光伏电站主动调节无功补偿策略[J]. 湖南电力, 2025, 45 (6): 84- 90.
	YU Xiaodong, LING Xu, WANG Yuan, et al. A reactive power compensation strategy with active regulation for photovoltaic power plants based on the improved holomorphic embedding method[J]. Hunan Electric Power, 2025, 45 (6): 84- 90.
8	杨胡萍, 龚家宁, 程明, 等. 计及多重不确定性的综合能源系统两阶段鲁棒低碳优化调度[J]. 中国电力, 2025, 58 (11): 101- 110, 121.
	YANG Huping, GONG Jianing, CHENG Ming, et al. Two-stage robust low-carbon optimal scheduling for integrated energy systems considering for multiple uncertainties[J]. Electric Power, 2025, 58 (11): 101- 110, 121.
9	陈水耀, 徐峰, 潘武略, 等. 适应高比例光伏接入的有源配电网自适应电流差动保护[J]. 浙江电力, 2025, 44 (1): 34- 43.
	CHEN Shuiyao, XU Feng, PAN Wulue, et al. Adaptive current differential protection for active distribution networks accommodating high levels of PV integration[J]. Zhejiang Electric Power, 2025, 44 (1): 34- 43.
10	殷林飞, 张依玲. 基于多重卷积组合大模型的光伏出力预测[J]. 综合智慧能源, 2025, 47 (4): 63- 72.
	YIN Linfei, ZHANG Yiling. Photovoltaic output prediction based on multi-convolutional combined large model[J]. Integrated Intelligent Energy, 2025, 47 (4): 63- 72.
11	吴晓刚, 阎洁, 葛畅, 等. 基于改进GRU-CNN的风光水一体化超短期功率预测方法[J]. 中国电力, 2023, 56 (9): 178- 186, 205.
	WU Xiaogang, YAN Jie, GE Chang, et al. Ultra-short-term power forecasting method for wind-solar-hydro integration based on improved GRU-CNN[J]. Electric Power, 2023, 56 (9): 178- 186, 205.
12	谭洪, 张雅丽, 关苏敏, 等. 含光伏光热制氢设备的园区综合能源系统低碳经济调度[J]. 浙江电力, 2025, 44 (7): 44- 54.
	TAN Hong, ZHANG Yali, GUAN Sumin, et al. Low-carbon economic dispatch of a park integrated energy system with PV/T-driven hydrogen production equipment[J]. Zhejiang Electric Power, 2025, 44 (7): 44- 54.
13	周洋, 黄德志, 李培栋, 等. 考虑平衡端点相位不对称及光伏接入的低压配电网三相潮流模型[J]. 中国电力, 2024, 57 (10): 190- 198.
	ZHOU Yang, HUANG Dezhi, LI Peidong, et al. A three-phase power flow model for low-voltage distribution networks considering balanced bus phase asymmetry and photovoltaic access[J]. Electric Power, 2024, 57 (10): 190- 198.
14	梁宏涛, 王莹, 刘国柱, 等. 光伏出力预测理论与方法综述[J]. 青岛科技大学学报(自然科学版), 2024, 45 (2): 147- 158.
	LIANG Hongtao, WANG Ying, LIU Guozhu, et al. Review on the theory and methods of photovoltaic output forecasting[J]. Journal of Qingdao University of Science and Technology (Natural Science Edition), 2024, 45 (2): 147- 158.
15	战文华, 车建峰, 王勃, 等. 基于网格化数值天气预报的区域光伏发电多输出功率预测方法[J]. 中国电力, 2024, 57 (3): 144- 151.
	ZHAN Wenhua, CHE Jianfeng, WANG Bo, et al. A grid-based numerical weather prediction method for multi-output prediction of regional photovoltaic power[J]. Electric Power, 2024, 57 (3): 144- 151.
16	张俊蔚. 基于物理模型的光伏电站输出功率预测[J]. 甘肃水利水电技术, 2015, 51 (1): 46- 49.
17	岳建通, 吴迪, 钱金跃. 基于ARIMA模型的分布式光伏出力预测方法应用[J]. 农村电工, 2024, 32 (2): 31- 32.
18	杨海亭, 白伟, 胡运冲. 分布式光伏发电接入智能电网功率预测模型优化研究[J]. 电工电气, 2024 (4): 1- 9.
	YANG Haiting, BAI Wei, HU Yunchong. Research on the optimization of power prediction model for distributed photovoltaic power generation connected to smart grid[J]. Electrotechnics Electric, 2024 (4): 1- 9.
19	付小标, 侯嘉琪, 李宝聚, 等. 一种二模态天气分型方法及其在光伏功率概率预测的应用[J]. 发电技术, 2024, 45 (2): 299- 311.
	FU Xiaobiao, HOU Jiaqi, LI Baoju, et al. A two-modal weather classification method and its application in photovoltaic power probability prediction[J]. Power Generation Technology, 2024, 45 (2): 299- 311.
20	王旭光, 李云, 王聪, 等. 基于超图像的短期光伏功率预测[J]. 广东电力, 2025, 38 (10): 14- 29.
	WANG Xuguang, LI Yun, WANG Cong, et al. Short-term photovoltaic power prediction based on hyperimage[J]. Guangdong Electric Power, 2025, 38 (10): 14- 29.
21	高海淑, 孙开宁, 黄钢, 等. 计及线路电热耦合特性的配电网鲁棒强化学习动态重构方法[J]. 电力系统自动化, 2026, 50 (1): 39- 50.
	GAO Haishu, SUN Kaining, HUANG Gang, et al. Robust reinforcement learning-based dynamic reconfiguration method for distribution networks considering line electro-thermal coupling characteristics[J]. Automation of Electric Power Systems, 2026, 50 (1): 39- 50.
22	殷豪, 李奕甸, 谢智锋, 等. 混合图神经网络和门控循环网络的短期光伏功率预测[J]. 太阳能学报, 2024, 45 (3): 523- 532.
	YIN Hao, LI Yidian, XIE Zhifeng, et al. Short-term photovoltaic power prediction method based on mixed graph neural network and gated recurrent unit network[J]. Acta Energiae Solaris Sinica, 2024, 45 (3): 523- 532.
23	王育飞, 郝德扬, 薛花, 等. 计及云图和混沌特性的光伏功率组合预测方法[J]. 太阳能学报, 2023, 44 (12): 74- 81.
	WANG Yufei, HAO Deyang, XUE Hua, et al. Combined forecasting approach of photovoltaic power based on cloud images and chaotic characteristics[J]. Acta Energiae Solaris Sinica, 2023, 44 (12): 74- 81.
24	邓韦斯, 戴仲覆, 王皓怀, 等. 基于WNN的光伏功率超短期预测研究[J]. 机械与电子, 2023, 41 (12): 15- 19.
	DENG Weisi, DAI Zhongfu, WANG Haohuai, et al. Ultra-short-term forecasting research research of photovoltaic power based on WNN[J]. Machinery & Electronics, 2023, 41 (12): 15- 19.
25	范国庆, 李康辉, 高捷, 等. 基于核密度估计和CatBoost算法的光伏功率预测方法[J]. 上海电力大学学报, 2023, 39 (6): 529- 535.
	FAN Guoqing, LI Kanghui, GAO Jie, et al. A photovoltaic power prediction method based on kernel density estimation and CatBoost method[J]. Journal of Shanghai University of Electric Power, 2023, 39 (6): 529- 535.
26	沙伟燕, 胡伟, 何宁辉, 等. 大规模虚拟储能平抑新能源功率预测误差优化调度方法[J]. 电力科学与技术学报, 2023, 38 (6): 167- 174.
	SHA Weiyan, HU Wei, HE Ninghui, et al. Optimal scheduling method for stabilizing power prediction error of new energy by large-scale virtual energy storage[J]. Journal of Electric Power Science and Technology, 2023, 38 (6): 167- 174.
27	卫志农, 徐昊, 陈胜, 等. 基于深度学习的直流配电网分布鲁棒优化调度方法[J]. 电力自动化设备, 2023, 43 (10): 87- 94.
	WEI Zhinong, XU Hao, CHEN Sheng, et al. Distributionally robust optimal dispatching method of DC distribution network based on deep learning[J]. Electric Power Automation Equipment, 2023, 43 (10): 87- 94.
28	杨锡运, 马文兵, 彭琰, 等. 基于组合神经网络的分布式光伏超短期功率预测方法[J]. 热力发电, 2023, 52 (8): 162- 171.
	YANG Xiyun, MA Wenbing, PENG Yan, et al. Distributed photovoltaic ultra-short-term power prediction method based on combined neural network[J]. Thermal Power Generation, 2023, 52 (8): 162- 171.
29	许伟欣, 杨明, 骆海琦, 等. 基于深度学习模型的光伏发电负荷预测[J]. 电气传动自动化, 2023, 45 (4): 62- 64, 49.
	XU Weixin, YANG Ming, LUO Haiqi, et al. Photovoltaic load forecasting based on deep learning model[J]. Electrical Drive Automation, 2023, 45 (4): 62- 64, 49.
30	朱明. 基于数据处理和深度学习的光伏发电预测模型[J]. 河南科学, 2023, 41 (7): 970- 977.
	ZHU Ming. Photovoltaic power generation prediction model based on data processing and deep learning[J]. Henan Science, 2023, 41 (7): 970- 977.
31	王海军, 居蓉蓉, 董颖华. 基于时空关联特征与B-LSTM模型的分布式光伏功率区间预测[J]. 中国电力, 2024, 57 (7): 74- 80.
	WANG Haijun, JU Rongrong, DONG Yinghua. Distributed photovoltaic power interval prediction based on spatio-temporal correlation feature and B-LSTM model[J]. Electric Power, 2024, 57 (7): 74- 80.
32	戚成飞, 王亚超, 李文文, 等. 基于数据驱动的高渗透率电动汽车充电规划与优化[J]. 中国电力, 2026, 59 (2): 104- 113.
	QI Chengfei, WANG Yachao, LI Wenwen, et al. Data driven planning and optimization of high penetration electric vehicle charging[J]. Electric Power, 2026, 59 (2): 104- 113.
33	高岩, 吴汉斌, 张纪欣, 等. 基于组合深度学习的光伏功率日前概率预测模型[J]. 中国电力, 2024, 57 (4): 100- 110.
	GAO Yan, WU Hanbin, ZHANG Jixin, et al. Day-ahead probabilistic prediction model for photovoltaic power based on combined deep learning[J]. Electric Power, 2024, 57 (4): 100- 110.
34	贺庆辰, 秦斌. 基于改进SAC算法的城轨列车混合储能系统动态功率分配策略[J]. 湖南电力, 2024, 44 (4): 11- 19.
	HE Qingchen, QIN Bin. Dynamic power allocation strategy for hybrid energy storage system of urban rail trains based on improved SAC algorithm[J]. Hunan Electric Power, 2024, 44 (4): 11- 19.
35	王斌, 明廷谦, 李志强, 等. 人工智能技术在配电网运行智能决策支持系统中的应用研究[J]. 电气技术与经济, 2024 (6): 116- 118, 122.

超参数	取值
块长度	8, 16
步长	8
注意力头数	4
编码器层数	3
训练轮次	30
批大小	16
早停耐心值	3
学习率	0.0001
输入序列长度	120
经验回放缓冲区大小	1000
训练数据子集数	3

超参数	取值
块长度	8, 16
步长	8
注意力头数	4
编码器层数	3
训练轮次	30
批大小	16
早停耐心值	3
学习率	0.0001
输入序列长度	120
经验回放缓冲区大小	1000
训练数据子集数	3

模型	学习率	编码器层数
Transformer	0.0001	3
Informer	0.0002	2
Autoformer	0.0001	2
PatchTST	0.0001	3
PTER	0.0001	3

模型	学习率	编码器层数
Transformer	0.0001	3
Informer	0.0002	2
Autoformer	0.0001	2
PatchTST	0.0001	3
PTER	0.0001	3

模型	E_MA	E_RMS	R²	C_or
Transformer	0.118	0.156	0.823	0.907
Informer	0.105	0.138	0.857	0.925
Autoformer	0.098	0.129	0.872	0.934
PatchTST	0.092	0.121	0.885	0.941
PTER	0.082	0.105	0.913	0.956