Smart DC microgrid fault protection

Direct current (DC) microgrids are power distribution systems that enable the integration of different energy sources, including fuel cells and photovoltaic arrays. Aside from being compatible with sustainable energy, they reduce the need for power conversions and provide other advantages in power transmission capacity, quality, and efficiency. However, standard protection measures fail to identify and accurately respond to DC faults, posing safety risks.

To facilitate the widespread use of DC microgrids, Anindita Deb and Arvind Kumar Jain developed a tri-layered deep learning-based framework that protects Standalone Low Voltage DC Microgrid Systems (LVDCMG) — commonly used in everyday settings — against faults.

“Our paper presents a reliable pathway toward AI-driven protection in LVDCMG, encouraging a shift from heuristic methods to data-centric solutions,” said author Anindita Deb. “It establishes a strong point for adopting noise-robust deep learning-based fault analysis, especially for remote or renewable installations where measurement disturbances are common.”

Their deep learning framework is composed of sequential layers that perform fault detection, fault type classification, and fault distance prediction. The first two layers are based on a Long Short-Term Memory (LSTM) algorithm designed for multitask learning and data storage and retrieval, much like the human brain. The final layer is a bidirectional LSTM that promotes feature fusion and predicts future outcomes.

Through simulation and sensitivity analysis, the authors found that their proposed model effectively learns from data patterns and adapts to different system topologies without prior knowledge. They hope similar deep learning approaches will be integrated into power distribution systems.

“The study can serve as a benchmark for researchers working on AI-assisted protection, encouraging more holistic evaluations that include noise, dynamic loading, and topology variations,” said Deb.

Source: “Design of an LSTM and Bi-LSTM driven intelligent protection scheme for standalone low-voltage DC microgrid,” by Anindita Deb and Arvind Kumar Jain, Journal of Renewable and Sustainable Energy (2025). The article can be accessed at https://doi.org/10.1063/5.0290701 .