1. School of Electrical and Electronic Engineering, Wenzhou University, Wenzhou, Zhejiang 325035; 2. Technology Institute of Wenzhou University in Yueqing, Yueqing, Zhejiang 325600
Abstract：Non-intrusive load monitoring (NILM), as the main way to analyze user's electricity consumption behavior, is of great significance to carry out energy consumption monitoring and realize electricity safety assessment. Aiming at the problem that the identification accuracy of the original voltage-current (V-I) trajectory features is not high, a recognition method based on the color coding of V-I trajectory features and the fusion of load high-order harmonic features is proposed in this paper. First, the high-frequency sampling data is preprocessed to extract the V-I trajectory and high-order harmonic characteristics of the load, and the numerical characteristics of instantaneous reactive power, power factor and current sequence distribution are mapped to the three channel pixel matrix of RGB color image by using color coding technology. Then, the high-order harmonic features are introduced and fused with three channel pixel matrix to form a mixed color image. Finally, the transfer learning of AlexNet network is used to train and classify the loads, which is verified by the PLAID data set and the measured data. The identification accuracy of the proposed load identification method is more than 95%, and the model has good generalization ability, which can be used for electricity safety management in home and similar occasions.
2022, Vol. 23 
