Time-Frequency Analysis (TFA) Method for Load Identification on Non-Intrusive Load Monitoring

Abstract

Public awareness of energy conservation can be realized through the process of electricity monitoring so that information can be found related to electricity usage and potential savings that can be made by the community. With the information on the results of this monitoring, the community can manage electricity usage more optimally so that it will have a positive impact on reducing electricity usage costs. The diversity of both linear and non-linear electrical equipment is a challenge in this monitoring process. In this study, the monitoring process is carried out in the aggregate on several electrical equipment indirectly on the electrical panel using the Non-Intrusive Load Monitoring (NILM) approach. The NILM approach is carried out through time and frequency domain analysis (TFA) to identify electrical equipment. Current signals (I) obtained in aggregate will be identified based on the characteristics of the frequency spectrum and its harmonics. The development of the TFA method was carried out to obtain feature vectors used for the identification process. Besides that, the noise reduction method was developed using the hybrid filter method by combining Median Filter and Average FFT filter (MFAFFT). The use of hybrid filter is done to eliminate noise through the frequency domain that was previously still remaining when filtered using Median Filter in the time domain. Classification performance of the test shows better accuracy results on hybrid filtered and using time and frequency feature. In SNR scenarios >15, the accuracy results are quite good with values greater than 0.9 and relatively stable at values above 0.9 at SNR >25.