PP: Data-driven classification of residential energy consumption patterns by means of functional connectivity networks

Purpose

Implement a good user aggregation and classification. 

or to assess the interrelation patterns between user profiles. 

Data

i. daily temperature and load profiles sampled at hourly for one year.  

ii. 2201 customers

iii.

Methodology

1. correlation matrix R of all pairs of users; 

2. transform the correlation matrix R into a distance matrix D;

?? why?因为相似度和距离是反的, 相似度越大,距离越近, 显然以距离作为边权重更合适.

3. network construction;

each user is a node; the weighted connections are established between all of them; weight eij = Dij.  

4. extract the MST network

Kruskal's algorithm

5. community detection

Newman. 

 Results

1. raw data: 各种distribution; average curve. 

2. distance matrix: 热度图; 

3. peak time?? 

4. 缺少与其他方法的比较,也没有说明具体分为了多少类,每个类的实际time series pattern是啥.反正乱七八糟的,生气.

你可能感兴趣的:(PP: Data-driven classification of residential energy consumption patterns by means of functional connectivity networks)