Classification of power quality disturbances using Kolmogorov-Arnold Network

The uninterrupted advancement and extensive implementation of renewable energy generation technologies have not only facilitated the optimisation of the energy structure and enhanced environmental benefits, but have also engendered novel challenges pertaining to power quality. The intermittency and uncertainty of new energy generation, such as wind energy, introduces additional power quality issues to the grid, including voltage fluctuations and harmonic pollution. Therefore, accurate classification of disturbances is essential for the identification of specific power quality issues arising from the integration of new energy sources into the grid. This enables the implementation of targeted management strategies to ensure the stable operation of the power system. In this paper, we propose a power quality interference classification method CNN-KAN based on convolutional neural network (CNN) and Kolmogorov-Arnold Network (KAN). The proposed method combines the automatic feature extraction capability of CNN with the advantages of KAN in key information capture or specific task processing. 12 synthetic power quality disturbances (PQDs) datasets, generated by an IEEE-1159 mathematical model, were used as training and test datasets. The results demonstrate the superiority of the proposed method, as evidenced by comparison experiments with two different methods.