Research Objective
To generate fault datasets from an inverter-dominated IEEE-33 bus AC microgrid and train AI models that detect fault occurrence, classify fault type and estimate fault location accurately.
System Architecture
The study includes IEEE-33 bus microgrid feeders, PV/wind inverter sources, BESS, grid-forming/grid-following converters, loads, breakers, fault events and measured phase/sequence signals for ML feature extraction.
Simulation Methodology
PowerFactory automation creates multiple fault types, locations, resistances and operating modes. The exported voltage, current, sequence and frequency features are used to train adaptive ML classifiers and location estimators.
Validation Scenarios
- LG, LL, LLG, LLL and high-impedance fault datasets
- Fault-resistance and location sweep across IEEE-33 feeders
- Grid-connected versus islanded microgrid cases
- Training/testing split with ML classification metrics
- Fault-location error and relay coordination comparison
Expected Graphs and Result Discussion
A complete result section should include the main waveforms, controller response, operating status and comparison tables needed for engineering thesis documentation. For this project, the important graph set includes:
- Three-phase voltage and current during faults
- Sequence current feature curves
- Fault-type confusion matrix
- Fault-location error plot
- Breaker status and relay response timeline
Thesis and Research Extension Ideas
XGBoost, Random Forest, CNN, LSTM, hybrid feature selection, noise robustness and explainable AI can turn the simulation into a publishable protection research workflow.