Fault Classification Using Hybrid FCN-BiLSTM Model in Grid Integrated with Distributed Generators

The grid integrated with distributed energy resources or generators (DERs) called microgrids are being used extensively because of power hunger in the world. Since the integration of more distributed generations, power systems have become more complex, and their protection is necessary for the reliable and continuous power supply. Accurate and highly precise fault detection and classification of fault play the most important role in complex power system protection. Conventional fault analysis techniques are not suitable for modern power systems because of bidirectional power flow in the power system due to integration for distributed renewable sources. This paper proposed an end to end deep learning-based hybrid Fully Convolutional Network (FCN) and Bidirectional Long Long-Term Memory (BiLSTM) model for fault classification tasks. Compared to machine learning, which requires preprocessing and human involvement, this automated model produces high accuracy and improved reliability. This proposed method is tested on various fault-simulated data in a grid integrated with two distributed generators. One distributed generator is a Diesel generator, and the other one is a Doubly Fed Induction Generator (DFIG) wind turbine. In this paper, we deploy the trained FCN-BiLSTM model in the Raspberry Pi microcontroller and give the simulated data for fault classification. For fault classification, 99.5% accuracy was achieved doing 10-fold cross validation using our proposed model. In the future work, real-time data from the OPAL-RT simulator can be fed to the Raspberry Pi microcontroller deployed with the proposed deep learning model for fault classification in complex power systems.