针对目前分布式光伏多机等值建模存在的建模复杂与精度低等问题,该文提出一种融合改进K-均值和卷积神经网络-长短时记忆神经网络（CNN-LSTM）的分布式光伏集群智能划分与等值建模方法。首先,选取包括输出电压、输出电流、并网点有功及无功功率,以及逆变器外环控制参数在内的多维度特征作为分群依据,通过引入手肘法确定分布式光伏集群的最优聚类数目。采用一种利用动态时间弯曲距离计算距离矩阵优化初始聚类中心的K-均值聚类算法,实现了对分布式光伏集群的智能划分,有效降低了分群的主观性与复杂性。该文结合卷积神经网络（CNN）与长短期记忆网络（LSTM）的优势,CNN用于捕捉空间维度特征,而LSTM则提取时间序列数据的动态特性,从而构建一个既能反映空间分布又能精准预测时间演变的CNN-LSTM分布式光伏集群等效模型。最后在PSCAD中仿真对比聚类等值模型与该文方法的分布式光伏集群等效模型的精度,该文采用模型在3种不同扰动下提升有功功率等值精度1.30、0.32与5.89个百分点,提升无功功率等值精度0.85、0.80与4.91个百分点,验证了该方法在提高分布式光伏系统建模效率与精度方面的有效性与实用性,为大规模分布式光伏并网系统的分析与优化提供了技术参考。
Abstract
The distributed nature and uncertainty of distributed photovoltaic（PV） systems pose new challenges to their equivalent modeling methods. In response to the existing issues of complex modeling and low accuracy in multi-machine equivalent modeling of distributed PV, this paper proposes an intelligent partitioning and equivalent modeling method for distributed PV clusters that integrates improved K-means and CNN-LSTM. Firstly, multi-dimensional features including output voltage, output current, active power and reactive power at the point of interconnection, and outer-loop control parameters of inverters are selected as the basis for clustering. The optimal number of clusters for distributed PV clusters is determined by introducing the elbow method. A K-means clustering algorithm that uses dynamic time warping distance to calculate the distance matrix and optimize the initial cluster centers is adopted to achieve intelligent partitioning of distributed PV clusters, effectively reducing the subjectivity and complexity of clustering. This paper combines the advantages of CNN and LSTM, where CNN is used to capture the spatial dimension features in distributed PV data, and LSTM focuses on extracting the dynamic characteristics of time series data, thereby constructing a CNN-LSTM equivalent model for distributed PV clusters that can reflect spatial distribution and accurately predict temporal evolution. Finally, the accuracy of the equivalent model of the clustering method and the distributed PV cluster equivalent model proposed in this paper is compared through simulation in PSCAD. The proposed model improves the equivalent accuracy of active power by 1.3, 0.32, and 5.89 percentage points, and the equivalent accuracy of reactive power by 0.85, 0.80, and 4.91 percentage points under three different disturbances, verifying the effectiveness and practicality of this method in improving the modeling efficiency and accuracy of distributed PV systems. This provides a technical reference for the analysis and optimization of large-scale distributed PV grid-connected systems.
关键词
光伏系统 /
聚类分析 /
神经网络 /
等值建模 /
K-均值聚类 /
手肘法
Key words
photovoltaic system /
cluster analysis /
neural networks /
equivalent modeling /
K-means clustering /
elbow method
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 冯玉军. 国际能源大变局下的中国能源安全[J]. 国际经济评论, 2023(1): 38-52, 4-5.
Feng Y J. China’s energy security under the drastic changes of international energy system[J]. International Economic Review, 2023(1): 38-52, 4-5.
[2] 张大海, 贠韫韵, 王小君, 等. 计及风光不确定性的新能源虚拟电厂多时间尺度优化调度[J]. 太阳能学报, 2022, 43(11): 529-537.
Zhang D H, Yun Y Y, Wang X J, et al.Multi-time scale of new energy scheduling optimization for virtual power plant considering uncertainty of wind power and photovoltaic power[J]. Acta Energiae Solaris Sinica, 2022, 43(11): 529-537.
[3] 习近平. 在第七十五届联合国大会一般性辩论上的讲话[J]. 中华人民共和国国务院公报, 2020(28): 5-7.
Xi J P.Statement at the general debate of the 75th session of the United Nations general assembly[J]. Gazette of the State Council of the People’s Republic of China, 2020(28): 5-7.
[4] 国家能源局. 我国风电光伏累计装机首超18亿千瓦[EB/OL].[2026-02-23].https://www.nea.gov.cn/20260130/503483f133e44016a66ddd6fad8ab695/c.html.
National Energy Administration.China's cumulative installed capacity of wind power and photovoltaic power exceeds 1.8 billion kilowatts for the first time[EB/OL]. [2026-02-23]. https://www.nea.gov.cn/20260130/503483f133e44016a66ddd6fad8ab695/c.html
[5] 韩平平, 林子豪, 夏雨, 等. 大型光伏电站等值建模综述[J]. 电力系统及其自动化学报, 2019, 31(4): 39-47.
Han P P, Lin Z H, Xia Y, et al.Review on equivalent modeling of large-scale photovoltaic power plant[J]. Proceedings of the CSU-EPSA, 2019, 31(4): 39-47.
[6] 关逸飞, 王玥娇, 刘奕元, 等. 计及资源差异特性的分布式光伏与储能聚合运营优化方法[J]. 山东电力技术, 2024, 51(12): 34-43.
Guan Y F, Wang Y J, Liu Y Y, et al.Distributed photovoltaic and energy storage aggregation operation optimization method for resource differences[J]. Shandong Electric Power, 2024, 51(12): 34-43.
[7] Chao P P, Li W X, Peng S M, et al.Fault ride-through behaviors correction-based single-unit equivalent method for large photovoltaic power plants[J]. IEEE Transactions on Sustainable Energy, 2021, 12(1): 715-726.
[8] 吕文琪, 刘子文, 董晓霄, 等. 考虑灵活性资源可调容量和响应速度的集群划分方法[J]. 电力工程技术, 2024, 43(5): 37-47.
Lyu W Q, Liu Z W, Dong X X, et al.A cluster partitioning method considering adjustable capacity and response speed of flexible resources[J]. Electric Power Engineering Technology, 2024, 43(5): 37-47.
[9] 潘学萍, 张源, 鞠平, 等. 太阳能光伏电站等效建模[J]. 电网技术, 2015, 39(5): 1173-1178.
Pan X P, Zhang Y, Ju P, et al.Equivalent modeling for photovoltaic power station[J]. Power System Technology, 2015, 39(5): 1173-1178.
[10] Li F, Wang Y C, Wu F, et al.Review of real-time simulation of power electronics[J]. Journal of Modern Power Systems and Clean Energy, 2020, 8(4): 796-808.
[11] 吴峰, 戴小妹. 计及高渗透率光伏的配电网广义负荷一体化等值建模[J]. 电力自动化设备, 2020, 40(5): 1-9, 46.
Wu F, Dai X M.Integrated equivalent modeling for generalized loads of distribution network considering high-penetration photovoltaic[J]. Electric Power Automation Equipment, 2020, 40(5): 1-9, 46.
[12] Du W J, Dong W K, Wang H F, et al.Dynamic aggregation of same wind turbine generators in parallel connection for studying oscillation stability of a wind farm[J]. IEEE Transactions on Power Systems, 2019, 34(6): 4694-4705.
[13] Chao P P, Li W X, Liang X D, et al.An analytical two-machine equivalent method of DFIG-based wind power plants considering complete FRT processes[J]. IEEE Transactions on Power Systems, 2021, 36(4): 3657-3667.
[14] Wang L, Yuan M Y, Zhang F, et al.Risk assessment of distribution networks integrating large-scale distributed photovoltaics[J]. IEEE Access, 2019, 7: 59653-59664.
[15] 叶林, 施媛媛, 王启亨, 等. 面向暂态分析的分布式光伏集群多步分群与等值建模[J]. 电力系统自动化, 2023, 47(14): 72-81.
Ye L, Shi Y Y, Wang Q H, et al.Multi-step grouping and equivalent modeling of distributed photovoltaic clusters for transient analysis[J]. Automation of Electric Power Systems, 2023, 47(14): 72-81.
[16] 潘学萍, 王卫康, 黄桦, 等. 考虑机网动态交互作用的光伏场群等值建模[J]. 电力自动化设备, 2023, 43(3): 80-85, 109.
Pan X P, Wang W K, Huang H, et al.Equivalent modeling of PV station groups considering dynamic interaction between PV and power grid[J]. Electric Power Automation Equipment, 2023, 43(3): 80-85, 109.
[17] Han P P, Lin Z H, Zhang J J, et al.Equivalent modeling of photovoltaic power plant based on factor analysis and correlation clustering[J]. IEEE Access, 2019, 7: 56935-56946.
[18] Wu H B, Zhang J J, Luo C, et al.Equivalent modeling of photovoltaic power station based on canopy-FCM clustering algorithm[J]. IEEE Access, 2019, 7: 102911-102920.
[19] Zhang Z, Liao S Y, Sun Y Z, et al.Unified dynamic equivalent model for distributed photovoltaic generation systems with different fault-ride-through strategies[J]. IEEE Transactions on Sustainable Energy, 2023, 14(4): 2062-2078.
[20] Øgaard M B, Frimannslund I, Riise H N, et al.Snow loss modeling for roof mounted photovoltaic systems: improving the Marion snow loss model[J]. IEEE Journal of Photovoltaics, 2022, 12(4): 1005-1013.
[21] 刘可真, 陈雪鸥, 陈镭丹, 等. 光伏发电系统动态离散等值模型研究[J]. 上海交通大学学报, 2023, 57(4): 412-421, 505.
Liu K Z, Chen X O, Chen L D, et al.Dynamic discrete equivalent model of photovoltaic power generation system[J]. Journal of Shanghai Jiao Tong University, 2023, 57(4): 412-421, 505.
[22] 蒋春容, 朱博汶, 周旭峰, 等. 基于分布式光伏集群控制的主动配电网电压优化策略[J]. 电力工程技术, 2025(3): 188-200.
Jiang C R, Zhu B W, Zhou X F, et al.Voltage optimization strategy foractivedistributionnetwork based ondistributed photovoltaic cluster control[J]. Electric Power Engineering Technology, 2025(3): 188-200.
[23] 李青. 基于注意力时间卷积神经网络的光伏功率概率预测[J]. 太阳能学报, 2025, 46(2): 326-332.
Li Q.Photovoltaic power probability prediction based on attention time convolutional neural network[J]. Acta Energiae Solaris Sinica, 2025, 46(2): 326-332.
[24] 彭曙蓉, 陈慧霞, 孙万通, 等. 基于改进LSTM的光伏发电功率预测方法研究[J]. 太阳能学报, 2024, 45(11): 296-302.
Peng S R, Chen H X, Sun W T, et al.Research on photovoitaic power prediction method based on improved LSTM[J]. Acta Energiae Solaris Sinica, 2024, 45(11): 296-302.
[25] 高任龙, 李峰, 马天东, 等. 融合多元时间序列与TCN-BiLSTM的分布式光伏功率超短期预测研究[J]. 山东电力技术, 2025, 52(12): 38-49.
Gao R L, Li F, Ma T D, et al.The research on ultra-short-term prediction of distributed photovoltaic power based on multivariate time series fusion with TCN-BiLSTM network[J]. Shandong Electric Power, 2025, 52(12): 38-49.
[26] 王钦, 姚伟, 艾小猛, 等. 含大规模面板级直流优化器的分布式光伏并网系统动态建模[J]. 电网技术, 2021, 45(2): 559-569.
Wang Q, Yao W, Ai X M, et al.Dynamic modelling of distributed PV grid-connected system with large scale of panel level DC optimizers[J]. Power System Technology, 2021, 45(2): 559-569.
[27] 盛万兴, 季宇, 吴鸣, 等. 基于改进模糊C均值聚类算法的区域集中式光伏发电系统动态分群建模[J]. 电网技术, 2017, 41(10): 3284-3291.
Sheng W X, Ji Y, Wu M, et al.Dynamic clustering modeling of regional centralized photovoltaic power plant based on improved fuzzy C-means clustering algorithm[J]. Power System Technology, 2017, 41(10): 3284-3291.
基金
江苏省自然科学基金青年项目（BK20241481）; 中央高校基本科研业务费专项资金资助（30922010709)
PDF(1996 KB)
