An Advanced Local Current-Only Protection for Microgrids: Deep-Learning-Based Approach

The integrity of protection actions in microgrids (MGs) is often compromised by varying operational conditions and contingencies. Notably, the operation of on-load tap-changers (OLTCs) can lead to the maloperation of overcurrent relays during both internal and external faults due to the altered fault current levels. This issue, combined with diverse operational conditions and other N-1 contingencies, further broadens the scope of undesirable operations. A deep-learning (DL) assisted fault detection method is introduced in this paper, which addresses the impact of OLTC operation on overcurrent protection in MGs under different operational scenarios. The proposed method utilizes a long short-term memory (LSTM) model to detect faults, taking into account the OLTC operation, on/off-grid status of the MG, outages of distributed generations (DGs), and capacitor bank switching. Additionally, various fault locations and resistances are incorporated into all simulated fault scenarios to enhance the training performance of the method. The ability of the LSTM model to interpret temporal dependencies in time-series signals allows the proposed method to rely solely on single-sourced current measurements at the relaying point, thereby reducing costs and improving module security. Simulation results on the low-voltage section of the IEEE 14-bus test system demonstrate the effectiveness of the proposed method in fault detection compared to existing approaches.