Smart-Meter Non-Intrusive Load Monitoring using Low Complexity Filter to Reduce Transient Spike

Optimal energy management requires the implementation of Non-Intrusive Load Monitoring (NILM) for detailed supervision of electrical energy loads. The information obtained from NILM on the use and power consumption of each equipment is useful for smart homes and demand response. However, some problems are still associated with implementing NILM, such as the transient spike, which is the difference between the initial and the steady state of home appliances with the current readings. The effect of the transient spike when monitoring electrical loads can be reduced using a Low Complexity Filter (LCF) with a median filter (MF). Based on simulation results with REDD and REFIT datasets for home appliances, MF can reduce the transient spike for all loads. This is similar to the experimental test result with a blender load using an MF71-window. $\Delta$I is stable, meaning that the MF can reduce the transient spike. The experimental test on this smart meter shows that the average accuracy is 0.925 from the average RE, PR, and F1 of 0.9-0.947. These results prove that the proposed method in this study can work optimally and recognize home appliances adequately.