Abstract: In order to apply the smart meters with load identification technology into real application，a nonintrusive load identification method using Mean-shift clustering algorithm with siamese network framework was

proposed. In this method，the load event was carried out for extracting the features，including active power and reactive poaer. Meanwhile，working time and working duration of the load was obtained to assist the load

identification. Secondly，Mean-shift clustering was used to classify the load events into the load type，and the load

type was then identified by comparing the siamese network with the load signature database. Thus，the load type

which the load event belongs to can be obtained. Finally，through experiments on the load test data of actual

household users，the result proved that the method proposed could obtain the load kinds within the user's home，and distinguish the load with a large coupling degree of power signatures which could provide the basis for application in smart meter.

      Key words: non-intrusive；time signature；smart meter；Mean-shift clustering；siamese network