Non-intrusive industrial load analysis utilizing bi-level clustering and improved fireworks algorithm

The surge in electricity demand has necessitated the development of efficient methodologies for analyzing overall electrical loads, through Non-Intrusive Load Monitoring (NILM) method. In this paper, we introduce a novel non-intrusive industrial load analysis method that adopts bi-level clustering with an enhanced fireworks algorithm. The methodology commences with the extraction of features from various operational states of industrial loads, culminating in the construction of a comprehensive feature library. Subsequently, a load decomposition model is formulated based on these features, and its optimization is achieved via the improved fireworks algorithm. Our experimental results robustly demonstrate that this approach substantially elevates load decomposition accuracy, consistently achieving over 89% accuracy across each load. This advancement not only underscores the method’s efficacy but also reinforces the potential of NILM technologies in industrial load analysis.