Novel High-Frequency Transient Techniques in Non-Intrusive Load Monitoring Techniques

The steady-state power signatures have the characteristics of addition and variation when the loads are operated at a simultaneous time and the system voltage changes, respectively. Therefore, it is difficult to identify individual loads in a steady-state system. The Discrete Wavelet Transform (DWT) has the characteristics of processing unstable and transient signals, analyzing time-frequency domain signals, and analyzing load signal characteristics. This paper proposes the analysis of DWT as a feature extraction technique for load identification. The proposed methods can improve the accuracy of load identification, shorten the computation time, reduce the memory sizes occupied by data processing, and overcome the difficulty of using steady-state features for non-intrusive load monitoring (NILM). This paper develops high-frequency transient sampling technology and cooperates with machine learning algorithms to establish a high-accuracy NILM model.