An improved Non-intrusive load identification method for V-I trajectory based on amplitude to pixel value

Abstract

Non-intrusive load monitoring (NILM) obtains electrical signals such as voltage, current and active power of the total load by measuring at the household electricity inlet, and disintegrates the total power consumption into a single device. The V-I trajectory used in current research can no longer reflect the electrical characteristics of different devices well, and the effect is not ideal when identifying the same type of equipment. Therefore, in this paper, a method of load identification based on the V-I trajectory of voltage and current load imprints is proposed to convert its amplitude into pixel value. The overall voltage and current changes are used as the standard for load identification, and the characteristics of V-I trajectory are extracted according to the difference between current and voltage changes in the cycle. Different pixel values were set in different amplitude variation intervals, and the obtained three-channel V-I trajectory map was input into the convolution neural network with the added attention module for load identification. The results show that the proposed V-I trajectory method based on amplitude transformation has good performance on some self-tested real data sets.