Comprehensive Non-Intrusive Load Monitoring Process: Device Event Detection, Device Feature Extraction and Device Identification Using KNN, Random Forest and Decision Tree

In recent years, a number of new studies have emerged due to the increase in technological developments. With the development of technology, data analysis and artificial intelligence studies have been proposed and studied in many areas. Increasing electrical energy demand and energy management need are some of the topics studied. NILM (Non-Intrusive Load Monitoring) is one of the methods studied during the implementation of these systems. With this method, it is aimed to detect and classify electrical devices used in power-consuming residences according to their characteristics by monitoring them from a center. Continuous monitoring of electrical energy in residences with Nilm works will enable leakage current detection and, thanks to the integration of renewable energy, to operate in home island mode in case of interruption. Along with the monitoring of the devices used in this direction, the characteristic features of the devices will be determined by preventing the use of reactive power and classifying the devices. In this study, hardware and software designs of data collected from an electrical network are presented according to the Nilm algorithm. With the hardware and software designs made, the stages of the NILM algorithm such as event detection, feature extraction and load identification are carried out. In addition, the designed device control cards enable the loads to be activated or deactivated in certain scenarios.