Faulty Line Localization in IEEE 30 Bus System Using CNN-LSTM

Faulty line localization is one of the crucial tasks for fault diagnosis along with fault classification to improve the reliability and sensitivity of complex or multi-bus power system protection. With this, the entire power system can be protected from fault and an uninterrupted power supply to the consumer can be met by removing and diagnosing the faulty line only. This paper introduces a deep learning-based hybrid model having Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM) that has been used to localize the faulty line in a multi-bus electrical power system network. This model is fully automated and achieves high accuracy with better reliability because no pre-processing or manual intervention is involved as in the case of machine learning. The proposed scheme is tested on simulated fault data of a standard IEEE 30 bus system by considering various faults configurations. The proposed technique is validated using 10-fold cross-validation and by evaluating various classification performance metrics. The weighted average accuracy of 99.36% has been achieved from this proposed model for faulty line identification in the considered system. The strength of the proposed model is also judged by comparing it with simple CNN and LSTM models. In the future, this model can be tested for the multiple tasks of fault diagnosis in the different complex power system models or with the real-time data of any power system network.