Li Jian, Li Shuoyu, Zhao Wen, Li Jiajie, Zhang Ke, Jiang Zetao
Metrology Center, Guangdong Power Grid Co.,Ltd., Guangzhou, 511545, China.
Power Supply Service, Dongguan Power Supply Bureau, Dongguan, 523576, China.
Sci Rep. 2024 Aug 22;14(1):19554. doi: 10.1038/s41598-024-68366-y.
The long-term loss of distribution network in the process of distribution network development is caused by the backward management mode of distribution network. The traditional analysis and calculation methods of distribution network loss can not adapt to the current development environment of distribution network. To improve the accuracy of filling missing values in power load data, particle swarm optimization algorithm is proposed to optimize the clustering center of the clustering algorithm. Furthermore, the original isolated forest anomaly recognition algorithm can be used to detect outliers in the load data, and the coefficient of variation of the load data is used to improve the recognition accuracy of the algorithm. Finally, this paper introduces a breadth-first-based method for calculating line loss in the context of big data. An example is provided using the distribution network system of Yuxi City in Yunnan Province, and a simulation experiment is carried out. And the findings revealed that the error of the enhanced fuzzy C-mean clustering algorithm was on average - 6.35, with a standard deviation of 4.015 in the situation of partially missing data. The area under the characteristic curve of the improved isolated forest algorithm subjects in the case of the abnormal sample fuzzy situation was 0.8586, with the smallest decrease, based on the coefficient of variation, and through the refinement of the analysis, it was discovered that the feeder line loss rate is 7.62%. It is confirmed that the suggested technique can carry out distribution network line loss analysis fast and accurately and can serve as a guide for managing distribution network line loss.
配电网发展过程中的长期线损是由配电网落后的管理模式造成的。传统的配电网线损分析与计算方法已无法适应目前配电网的发展环境。为提高电力负荷数据中缺失值填充的准确性，提出采用粒子群优化算法优化聚类算法的聚类中心。此外，可利用原始的孤立森林异常识别算法检测负荷数据中的离群点，并采用负荷数据的变异系数提高算法的识别精度。最后，本文介绍了一种基于广度优先的大数据环境下线损计算方法。以云南省玉溪市的配电网系统为例进行了仿真实验。结果表明，在部分数据缺失的情况下，改进后的模糊C均值聚类算法的误差平均为 -6.35，标准差为4.015。在异常样本模糊情况下，改进后的孤立森林算法的特征曲线下面积为0.8586，基于变异系数的下降最小，经细化分析发现馈线线损率为7.62%。证实了所提技术能够快速、准确地进行配电网线损分析，可为配电网线损管理提供指导。
