ScholarWorks@경상국립대학교

초록

This paper presents a reliability evaluation of a dc Solid-State Circuit Breaker (dc-SSCB) equipped with a snubber circuit, based on the FIDES 2022 failure library and the Parts Stress Analysis (PSA) methodology, utilizing a Long Short-Term Memory (LSTM) neural network. To validate the performance, actual temperature data from Seoul, South Korea, was used. The system failure rates and reliability were analyzed using both the conventional failure rate method and LSTM-based predictions of operating temperature. Furthermore, the predicted temperatures were compared with actual measured values to assess the accuracy of the prediction model. The analysis results show that the LSTM-based approach predicts the system reliability to reach 0.5 after approximately 14 years of operation, which is about 12.5% lower than the result obtained using the conventional static method based on average failure rates. Moreover, when compared with the reliability evaluated using actual measured temperature data, the prediction error of the LSTM-based approach was found to be less than 1%, demonstrating high accuracy. These findings indicate that incorporating time-varying thermal stress information enables more precise and realistic reliability assessments that closely reflect actual operating conditions.

키워드