Моніторинг неінтрузивного навантаження: економічний підхід до ідентифікації побутової техніки з використанням машинного навчання

This research focuses on developing a cost-effective non-intrusive load monitoring system (NILM) to identify household appliances using machine learning, specifically the k-nearest neighbors (kNN) algorithm which is not disturbing the existing system. The object of this research is the process of appliance identification based on power consumption characteristics in residential energy monitoring. The main problem to be solved is the lack of accessible, affordable, and efficient tools for monitoring household electricity consumption, as existing solutions are often costly or require complex installations. Existing solutions are expensive or require complicated setup. This research seeks to design a low-cost NILM that can identify household appliances without invasive system while ensuring high accuracy. This study successfully designed and implemented an electrical recording device that integrates machine learning algorithms, achieving an identification accuracy of 83.33 % across six test scenarios involving various household appliances. The findings of this study show that utilizing active power and power factor as classification parameters allows for effective equipment identification. The moderate accuracy of the system indicates that the proposed design is quite promising but can be improved with more advanced algorithms and additional sensor data. The resulting system is cost-effective due to its inexpensive components which are achieved due to the modular design and the use of inexpensive components, such as the Wemos D1 mini and PZEM-004T V3 sensors, which simplify implementation and enhance system scalability. Its built-in LCD provides real-time monitoring without the need for internet connectivity. This research demonstrates the feasibility of a scalable and cost-effective NILM system, which can be further improved with advanced algorithms and additional sensor data for broader applications in smart energy management