Commutation Failure Boundary Detection Based on Deep Recurrent Neural Network

In hybrid AC and DC grid, commutation failures will likely cause protection malfunction and may finally lead to serious cascading failures. Therefore, the accurate commutation failure boundary detection is of great value to improve AC protections’ performance, optimize AC and DC protections’ coordination and ensure power grid’s safety and stability. A new approach for fast commutation failure diagnosis is proposed based on DRNN(Deep Recurrent Neural Network). By using the converter bus voltages, DC current and the trigger angle commands of the inverter, this method can consider multiple factors affecting the commutation process, can trace the cause of commutation failure accurately and has certain predictive function.