随着分布式能源在台区中占比不断提升和台区智能化发展不断深入，提出一种面向高渗透率分布式能源台区的改进的k-means用户用电行为分析方法。首先基于误差平方和(sum of squares errors，SSE)构造的最优k值策略和由动态加权思想优化密度参数的计算提出一种改进的密度预聚类算法，选取聚类初始簇和初始中心；然后融合欧氏距离和余弦相似度等构造多特征距离相似度优化k-means聚类；最后使用含高渗透率分布式能源台区实测数据进行分析，结果表明本文所提算法模型较对比模型可以更准确的识别用户所在簇，提取典型用户行为模式，在给定聚类指标上得到较好的提升。
英文摘要:
With the increasing proportion of distributed energy in the substation area and the deepening of intelligent development of the substation area, an improved k-means user electricity behavior analysis method for high penetration distributed energy substations was proposed. Firstly, based on the optimal k value strategy constructed by sum of squares errors (SSE) and the calculation of density parameters optimized by dynamic weighting idea, an improved density pre-clustering algorithm was proposed to select the initial cluster and initial center. Then, Euclidean distance and cosine similarity were combined to construct multi-feature distance similarity to optimize k-means clustering. Finally, the measured data of distributed energy substations with high permeability are used for analysis. The results show that the proposed algorithm model can identify the cluster of users more accurately than the comparison model, extract typical user behavior patterns, and get a better improvement on the given clustering index
