摘要
分布式光伏与数字孪生理念相结合,是应对大规模并网光伏群控挑战的有效方法,也是推进新型电力系统建设的重要内容。该文提出一种面向分布式光伏群调群控的数字孪生方法,拟从光伏一致性表征、孪生模型重建、功率推演预测3个方面支撑孪生系统构建。基于K-means算法提出光伏一致性表征方法,利用皮尔逊系数加权构造气象因子与光伏并网节点的电压灵敏度作为指标,对配电网电压影响相似的光伏进行聚类;以运动恢复结构(structure from motion,SFM)为原理提出基于加速稳健特征(speeded up robust features,SURF)算法的高质量相机位姿估计方法与低误差稀疏点云重建方法,再结合多视角立体视觉(multiple view stereo,MVS)原理提出点云稠密化方案,形成基于SFM-MVS的高保真光伏模型重建算法;基于长短期记忆网络(long-short term memory,LSTM)网络提出光伏短期功率预测方法,利用皮尔逊系数加权处理输入的光照、温度数据,并提出残差补偿机制和物理意义约束两部分提高预测结果准确性。文章最后通过实验,验证集群划分的合理性、光伏模型重建优越性以及功率预测的准确性。
The combination of distributed PV and the digital twin concept is an effective method to cope with the challenges of large-scale grid-connected PV group control and an important element in promoting the construction of new power systems.This paper proposes a digital twin method for distributed PV group regulation and group control.It intends to support the construction of the twin system from three aspects:PV consistency characterization,twin model reconstruction,and power deduction and prediction.A PV consistency characterization method is proposed based on the K-means algorithm,using Pearson's coefficient weighting to construct a meteorological factor with the voltage sensitivity of the PV grid-connected nodes as an indicator to cluster PVs with similar voltage impacts on the distribution network.A high-quality camera pose estimation method based on the SURF algorithm.A low error sparse point cloud reconstruction method is proposed based on the principle of SFM,and then a point cloud densification scheme is proposed in combination with the principle of MVS to form a high-fidelity PV model reconstruction algorithm based on SFM-MVS.A PV short-term power prediction method is proposed based on LSTM network,which utilizes Pearson's coefficient weighting to process the input light and temperature data,and proposes a residual compensation mechanism and physical significance constraints in two parts to improve the prediction result accuracy.The article concludes with experiments to verify the rationality of cluster division,the superiority of PV model reconstruction,and the power prediction accuracy.[基金项目:国家电网公司科技项目(No.5100-202113564A-0-5-SF)。Project Supported by the Science and Technology Project of Grid Corporation of China(5100-202113564A-0-5-SF).]
作者
蔡瑞天
姚丽娟
武昕
CAI Ruitian;YAO Lijuan;WU Xin(School of Electrical and Electronic Engineering,North China Electric Power University,Changping District,Beijing 102206,China)
出处
《电网技术》
北大核心
2025年第2期593-603,I0061-I0063,共14页
Power System Technology
基金
国家电网公司科技项目(No.5100-202113564A-0-5-SF)。
关键词
数字孪生
分布式光伏
三维重建
集群聚类
功率预测
digital twins
distributed photovoltaic
three-dimensional reconstruction
cluster clustering
power prediction
