CAJ | 학술논문

结合k-means、k-medoids、SOM以及FCM等聚类算法，构建了电力大客户典型用电模式的聚类分析模型，提出了一种评估聚类效果的新方法。首先通过分析电力客户用电指标数据及其特点，提出采用高斯滤波器对含“噪声”曲线数据进行平滑处理来获取客户用电数据。然后提出了聚类平均半径、平均直径和平均最小间距等3个评价指标，并以此为基础设计出一种评估聚类得分的新方法。最后使用聚类分析模型对某地区电力大客户日用电量曲线进行聚类分析，实现了地区典型用电模式的自动识别功能。实际算例分析结果表明，该评估方法物理概念清晰、简便、实用。
결합k-means、k-medoids、SOM이급FCM등취류산법，구건료전력대객호전형용전모식적취류분석모형，제출료일충평고취류효과적신방법。수선통과분석전력객호용전지표수거급기특점，제출채용고사려파기대함“조성”곡선수거진행평활처리래획취객호용전수거。연후제출료취류평균반경、평균직경화평균최소간거등3개평개지표，병이차위기출설계출일충평고취류득분적신방법。최후사용취류분석모형대모지구전력대객호일용전량곡선진행취류분석，실현료지구전형용전모식적자동식별공능。실제산례분석결과표명，해평고방법물리개념청석、간편、실용。
In order to gain the large power customers’ typical power consumption profiles in a power supply area, a new clustering evaluation method is presented and a clustering analysis framework based on k-means, k-medoids, self-organized maps (SOM) and Fuzzy C-Means (FCM) is built. It analyzes the characteristic of the electricity consumption data and uses the Gaussian smoothing method to reduce the noise in the data. Clusters average radius, clusters average diameter and clusters average minimum distance are proposed and used to design the clustering evaluation method. This framework is utilized to analyze the daily electricity consumption curves of the whole customers in a certain area, which can automatically recognize the number of clusters. The result shows this methodology is clear in physical conception, simple and practical.