GridGuard AI

GridGuard AI ⚡ Inspiration

Power failures can affect homes, businesses, hospitals, and many other places. We were inspired by the idea of preventing a power failure instead of reacting after it happens. We wanted to build a system that could identify early warning signs in electrical equipment and help people take action before a serious failure occurs.

What We Built

GridGuard AI monitors important transformer parameters such as load, temperature, voltage, and current. It uses machine learning to identify unusual patterns and estimate the risk of a possible failure. The system then explains the possible reason and suggests preventive actions.

We also added a What-If Simulation, which allows users to see how the risk could change when the transformer load is reduced.

How We Built It

We built the backend using Python and FastAPI, with machine learning for anomaly detection. The dashboard was developed using React, and we used simulated sensor data to represent real-time transformer conditions. We also designed the system to use AWS Strands Agents for AI-based reasoning and recommendations.

What We Learned

Through this project, we learned how to combine machine learning, AI agents, real-time data, and web development to solve a real-world problem. We also learned that AI should support human decision-making rather than blindly control critical systems.

Challenges

One of our main challenges was creating realistic sensor data without access to a real electrical grid. We also had to decide how to identify meaningful abnormal patterns and present the results in a simple way. Another challenge was designing the system with human approval for safety-critical decisions.

Our Goal

Our goal with GridGuard AI is simple:

Predict the problem before the blackout, understand why it may happen, and help people take preventive action.

Built With

• agents
• ai
• amazon-web-services
• artificial
• energy
• grid
• intelligence
• iot
• learning
• machine
• maintenance
• pandas-and-numpy-for-data-processing
• power
• predictive
• python
• python-and-fastapi-for-the-backend
• react
• scikit-learn-for-anomaly-detection
• strands
• sustainability