The research on low-voltage load simulation and evolution method for distribution substations based on nonintrusive load monitoring

Abstract Driven by the “Double Carbon” strategic goal and energy transformation, intelligent, and refined distribution networks are crucial for low-carbon power systems. Traditional load analysis methods fall short in handling diverse, low-power loads in distribution areas. This paper presents a comprehensive methodology utilizing nonintrusive load monitoring (NILM) technology, featuring an enhanced transformer-based load decomposition algorithm with multiscale feature extraction and dual-layer attention mechanisms. Additionally, we develop a load simulation model grounded in spatiotemporal correlations and a multifactor coupling evolution model for precise load prediction. Experimental results validate improvements in load decomposition accuracy, simulation fidelity, and evolution prediction reliability.