Short-circuit Fault Classification Method for AC Submarine Cables in Offshore Wind Farms Based on Improved Sparse Representation

Accurate and rapid submarine cable fault classification is an important part of the maintenance in offshore wind farms. This paper propose a short-circuit fault classification method for AC submarine cables in offshore wind farms based on improved sparse representation, which utilized time-domain characteristics of half-cycle current signals as the basis for fault classification. The proposed method applies a K-SVD dictionary learning algorithm to extract various types of fault signal features into the corresponding over-complete dictionaries. In the light of learning dictionaries, an improved sparse decomposition algorithm using mixed alternating direction method of multipliers (M-ADMM) is proposed to factorize the fault signal into the product of dictionary and sparse vector. Combined with the classification method based on sparse representation, the classification of the reconstructed fault signal is achieved. PSCAD/EMTDC simulation results indicated that the improved sparse decomposition algorithm had an excellent ability in signal reconstruction and noise reduction. The fault classification method do not need to manually design fault signal features. Compared with SVM, CNN, LSTM, it has a high classification accuracy and improved noise robustness in various operation scenarios. ©2023 Chin.Soc.for Elec.Eng.