**5. Summary**

The paper presented the system for NILM task based on the transient features of the generated pulse analysis. It exploits mutual influence between devices operating in the same power circuit to identify the moment of introducing the new one.

The pulse signal generator was designed. Its task is to generate current pulses at predetermined time intervals in a fixed phase of the supply voltage. The measurement system acquires current pulses and stores them as sample vectors. A method for processing current signals was designed to determine their characteristic features based on the crosscorrelation calculated for each pair of EAs. The method uses information about the phase and amplitude of all (periodic and non-periodic) components of the current pulse appearing in the transient state of the device turned on. The processing result is a signature with features characterizing the EA. The signature quality was verified using three different classifiers.

The presented experiments show that devices connected to one circuit of the supply network influence each other. A significant impact of a background device in the steady state on the current pulse on another device being turned on was observed. When a known load is switched on under repeated conditions, the change in the shape of this pulse may be characteristic for a device operating in the background. Results of the classification show that in the best case, 9 out of 15 EAs are recognizable with an accuracy of at least 97%. Satisfying results were obtained for majority of tested EAs. There are types of EA for which this method fails. Therefore multiple different identification methods should be implemented simultaneously.

In practical application of NILM system the changing set of devices operating at the same time must be considered. This makes the task difficult, as the change of the pulse shape will be a certain superposition of all working EA. Therefore additional research are required to approach this challenge. Results of presented experiments show that in the highly controlled environment (especially when only a single appliance is operating) the proposed approach provides high identification accuracy Its applicability should be further investigated in the future.

**Author Contributions:** Conceptualization, R.K. and R.Ł.; Data curation, A.W.; Formal analysis, R.K.; Funding acquisition, R.Ł. and P.B.; Investigation, A.W. and P.B.; Methodology, R.K. and P.B.; Resources, A.W. and R.Ł.; Visualization, A.W. and K.D.; Writing—original draft, A.W.; Writing— review & editing, A.W., R.K., P.B. and K.D. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author.

**Conflicts of Interest:** The authors declare no conflict of interest.
