Nonintrusive Industrial Load Monitoring Considering Load Power Characteristics and Timing Correlation

Implementing nonintrusive load monitoring (NILM) for industrial consumers plays a vital role in managing power demand and enhancing energy utilization efficiency. Focusing on the scarcity of sampling data and continuous variation of loads in industrial settings, this article proposes a nonintrusive industrial load decomposition (LD) method that considers the load power consumption characteristics and time series correlation. According to the time-varying characteristics of the power consumption, the load is divided into three types including switching load, multi state load, and continuously varying load. On this basis, the active and reactive power characteristics are jointly considered, and integer programming is used to build the load decomposition model. Particularly, the matrix factorization (MF) method is used to describe the continuously varying load. In addition, the timing correlation constraints under the base vector grouping constraint and production process constraints are proposed and integrated into the load decomposition model. Finally, the proposed method is validated on a public dataset using a PC platform and a Raspberry Pi 5, respectively. The results of the tests on the PC platform show that the proposed method has higher accuracy than the existing methods.