Research on ELM-Based Fault Diagnosis of AC Filter Circuit Breaker in Converter Station

In the high-voltage DC transmission system, the converter station is a site established to complete the conversion of AC and DC power. Whether the AC filter can operate stably will directly affect the reliability of the DC transmission system. At the same time, AC filter also have higher fault rate than other circuit breakers in the converter station. If the operating conditions of the AC filter can be predicted to a certain extent and hidden dangers can be removed in advance, serious accidents such as fault trips can be avoided to a large extent. This paper proposes a fault early warning method of AC filter circuit breaker based on the extreme learning machine (ELM). Taking a converter station in Northwest China as the background, analyze the recorded waveform of the AC filter. The opening and closing current of the circuit breaker and the corresponding operating conditions when the AC filter is switched on and off are extracted, establishing a learning training library and verifying its effectiveness. The influence of different activation functions and the number of hidden layer nodes on this method is discussed. Finally, the fault diagnosis performance of intelligent learning methods such as BP neural network, RBF neural network and genetic RBF network is compared.