A fault classification scheme based on protective agents for microgrid with parameters impact analysis.

Area of Science:

• Electrical Engineering
• Power Systems
• Artificial Intelligence

Background:

• Microgrid protection faces challenges in fault identification.
• Overcurrent protection is common, but deep neural networks (DNNs) offer improved accuracy.
• Deep learning-based fault detection and location (FDL) schemes can be sensitive to parameter variations.

Purpose of the Study:

• To present a protective agent (PA)-based fault classification method using intelligent electronic devices (IEDs) and hybrid DNNs for microgrids.
• To investigate the impact of various parameters on the performance of DNN-based FDL schemes.
• To conduct a comprehensive sensitivity analysis of the proposed FDL scheme.

Main Methods:

• Developed an FDL scheme utilizing two types of DNNs (single and hybrid layers) and PAs.
• Analyzed overcurrent-based protection scenarios and parameter impacts.
• Performed data and sensitivity analyses on factors like microgrid topology, IEDs, DNN structure, protection scheme, and data transfer.

Main Results:

• The proposed algorithm achieved high accuracy, with fault detection ranging from 95.54% to 99.96%.
• Fault type/phase detection accuracy was between 95.56% and 99.86%, with fault location error between 1.27% and 6.51%.
• Hybrid DNNs (DNN-2) consistently outperformed single-layer DNNs (DNN-1), though parameter sensitivity was observed.

Conclusions:

• The PA-based hybrid DNN approach enhances microgrid fault classification and location accuracy.
• Parameter selection and microgrid configuration significantly influence the performance of DNN-based protection schemes.
• The study provides valuable insights for optimizing microgrid protection using intelligent methods.