Clustering and Markov Model of Plug-In Electric Vehicle Charging Profiles for Peak Shaving Applications

Table 1

Comparison of relevant references.

References	Clustering methods	Characteristics
[4]	Clustering based on two-branch and three-way decisions	(1) Two-branch decisions and three-way decisions are incorporated for basic clustering selection.
		(2) The optimal threshold and time consumption are the problems.
[5]	Ensemble spectral clustering	(1) Clustering integration based on weighted consistency matrix of clustering evaluation index.
		(2) How to improve the construction method of the consistency matrix and how to fully explore the similarity information of user load characteristics have not been well solved.
[7]	Hierarchical clustering	(1) The results are shown by dendrogram.
		(2) The initial number of clusters is not required.
		(3) Clustering efficiency and clustering accuracy cannot be satisfied at the same time.
[8]	Density-based spatial clustering	(1) Sensitive to the input parameters.
		(2) The clustering convergence time is longer as the dataset is larger.
[10]	Mean shift clustering	(1) The methods can describe the profile’s characteristics from the overall or macroscopic level.
[11]	Information theoretical clustering	(2) The algorithms lose some important information about the profile’s shape pattern.
[14]	Gaussian mixture model clustering	(1) The probability that a given point belongs to each of the possible clusters is provided.
		(2) Complete data information for prediction is required. The validity in high-dimensional spaces is missing.
[15, 18, 19]	Double-layer clustering	(1) The methods can effectively identify the changing trend of the electricity consumption profiles.
[16, 17]	Segment aggregation approximation and spectral clustering	(2) The quality of the algorithm is greatly affected by the large variation in the data.