Protection in inverter-based microgrids is challenging because fault current magnitude and waveform characteristics change with operating mode, converter control and renewable penetration. AI-based adaptive protection allows the relay logic to use richer measurements instead of fixed current thresholds.
Research problem
Traditional overcurrent coordination can become unreliable when the same feeder operates grid-connected and islanded. Converter current limiting, changing fault levels and high-impedance faults make adaptive logic attractive.
Data and features
A strong study can use phase voltages and currents, sequence components, frequency, RoCoF, P/Q, apparent impedance and breaker states. Fault classes may include LG, LL, LLG, three-phase faults and high-impedance faults across multiple fault resistances and locations.
AI models and validation
Random Forest, XGBoost and sequence models such as LSTM can be compared using detection accuracy, classification accuracy, location error, inference time and robustness to unseen operating conditions.
Simulation architecture
PowerFactory or MATLAB/Simulink can generate labelled disturbance data. A Python pipeline can perform feature engineering, train models and return adaptive decisions for relay settings or breaker logic.
Need this topic implemented as a simulation project?
Share your research title, abstract or base paper. The model, controller, case studies and required plots can be scoped around your research objective and software version.
