The electrical distribution system relies on the qual-ity of insulation to ensure the power supply. When an insulator has a high level of surface contamination, the leakage current increases, leading to a disruptive discharge. To mitigate this problem, one solution is to evaluate the leakage current in relation to the insulator's contamination, as presented in the IEC 60507 standard. To improve the ability to predict faults in insulators, this paper proposes using the neural basis expansion analysis for time series forecasting (N-BEATS) to predict faults in the power grid. For signal denoising, the empirical mode decomposition (EMD) methods are applied. The model proposed in this paper has an input filter stage ensuring that only the trend in leakage current variation is evaluated. The proposed method shows promise, being superior to other deep learning models.

--

Deep learning,Fault prediction,Power system

• Outras Engenharias e Tecnologias - Engenharia e Tecnologia
• Engenharia Eletrotécnica, Eletrónica e Informática - Engenharia e Tecnologia