Know the fault before the crew rolls out.
GridGuard analyses voltage and current signals from transmission lines and classifies the fault type automatically, so operators can respond faster and restore power sooner.
The problem
Every minute a fault goes unidentified costs reliability.
Transmission lines carry power across long distances and are exposed to weather, equipment ageing and growing loads from renewables. When a fault hits, identifying its type quickly is the first step to isolating it and restoring supply.
of electricity grid lines operate worldwide.
accuracy achieved by our decision tree and random forest models.
per year in grid investment is needed by 2030 to keep pace with energy transitions.
Sources: IEA, Electricity Grids and Secure Energy Transitions (2023); accuracy from the founding research paper (IEEE, 2025).
How it works
Signals in. Fault type out.
GridGuard plugs into the measurements utilities already record at substations.
Measure
Three-phase voltage and current signals are captured from the line.
Balance
SMOTE and class weighting correct for rare fault types in the data.
Classify
Optimised ML and deep learning models identify the fault type.
Alert
Operators receive the fault class to guide isolation and repair.
Product
Engineered for reliable, fast classification.
Models are tuned with grid search and Bayesian optimisation (Optuna).
High-accuracy models
Decision tree and random forest models reached 97% accuracy in testing.
Balanced training
SMOTE raised CNN accuracy from 86.56% to 87.58% by addressing class imbalance.
Robust neural networks
Batch normalisation, dropout and cosine-annealing learning rates improve reliability.
Smart-grid ready
Roadmap includes renewable energy integration and smart-grid deployment.
Who it's for
For the teams that keep the lights on.
Transmission operators
Faster fault identification in control rooms.
Distribution utilities
Prioritise crew dispatch with clearer fault information.
Renewable developers
Monitor lines connecting solar and wind plants to the grid.
The research
Tested across machine learning and deep learning.
- Applies machine learning and deep learning to classify transmission line fault types.
- SMOTE and class weighting, optimised with grid search and Optuna, improve performance.
- Decision tree and random forest maintained 97% accuracy; CNN rose from 86.56% to 87.58% with SMOTE.
- Future work: larger datasets and adaptation to renewable energy and smart grids.
Mousa, A., Zraqou, J. S., Shaheen, H., & Zawahra, I. (2025). Fault Type Classification in Electrical Transmission Lines. IEEE ICCIAA 2025.
Team
Power engineering meets AI.
Founded in 2025 by a student team from the faculty below.
Faculty of Engineering & Information Technology
Power systems analysis, signal processing and machine learning.
FAQ
Questions we hear often.
What fault types can GridGuard identify?
It classifies the standard transmission line fault categories, such as line-to-ground, line-to-line and three-phase faults.
Does it replace protection relays?
No. It complements existing protection by giving operators a fast, data-driven fault classification.
What data is required?
Three-phase voltage and current measurements already recorded by most substations.
Can it run in real time?
Lightweight models like decision trees are well suited to fast inference in operational settings.
Make your grid smarter about faults.
We are looking for utilities and grid operators to pilot GridGuard on real line data.