A Building Electrical Fault Diagnosis Method Based on CEEMDAN and Neighborhood Rough Sets

The building electrical fault diagnosis methods based on machine learning often require the extraction of a large number of characteristic signals for fault diagnosis. However, it is challenging to obtain some valid feature signals from a large number of feature signals. To tackle the aforementioned challenges, we present a novel fault diagnosis strategy tailored for building electrical systems. This approach integrates neighborhood rough sets (NRS) with the comprehensive ensemble empirical mode decomposition featuring adaptive noise (CEEMDAN). The initial step involves utilizing the CEEMDAN algorithm to decompose the fault signal, enabling a detailed examination of its intricate characteristics. Then, NRS solves the problem that traditional rough sets require to discretize the data by adding the domain concept. The proposed method is tested on a dataset obtained from the building electrical experimental platform board MA2067. The results of the experiments indicate that the methodology we have proposed achieves an impressive fault diagnosis accuracy of 96.078%, thereby demonstrating its efficacy and precision.