Power value is a main feature in existing non-intrusive load monitoring(NILM)methods

but it is not suitable for low power appliances to meet the fine demand of intelligent electricity. The current waveform

power and harmonic feature of multiple appliances are firstly analyzed. The most distinct harmonic amplitude in the frequency domain

for the low power appliances

is chosen as a new feature. A new NILM method is presented

in which the delta feature extraction is used to get variations of load features

with information entropy adopted to determine optimal cluster number and load similarity by calculating inter-cluster entropy. Fuzzy clustering is also used to monitor the quantity and kind of appliances. Finally

experiment results have proved that the proposed method has higher accuracy and stability

and identification accuracy of low power appliances is improved observably. This work is supported by National High Technology Research and Development Program of China(863 Program)(No. SS2015AA050203)

State Grid Corporation of China and Fundamental Research Funds for the Central Universities(No. 2015XS05).