Non-Intrusive Load Monitoring via Binary Integer Search and Transient Load Feature Integration

Non-intrusive load monitoring (NILM) aims to analyze energy consumption and monitor loads without intrusive measurements. However, existing household NILM models often depend heavily on specific datasets or electricity usage scenarios, which limits their performance consistency across different households. To enhance adaptability across diverse households and regions, this paper proposes an optimization-based NILM solution consisting of two key steps: using binary integer search to determine a feasible solution space, and incorporating transient load features in the final solution selection. The proposed model only requires a few electrical parameters for each appliance, eliminating the need for device-level load training data. Tests on real-world datasets across multiple households demonstrate the models effectiveness in load disaggregation and its computational efficiency, making it suitable for practical applications. Furthermore, the models physical interpretability lays a strong foundation for further research and development.