-
-
Preventing battery failures before they happen
Tagline
Preventing battery failures before they happen, saving ₹3.6 Cr annually.
Inspiration
Battery Smart operates 1,500+ swap stations processing 125,000 daily swaps. We discovered that 2-3% of batteries fail unexpectedly each month, costing ₹8,000 per failure in replacement costs, lost revenue, and customer dissatisfaction. More critically, thermal runaway poses fire risks. We built SwapGuard to predict failures 7-14 days in advance.
What It Does
SwapGuard AI monitors real-time telemetry from 50,000+ batteries across Battery Smart's network and uses machine learning to predict which batteries will fail in the next 7-14 days. Station partners receive automated alerts via WhatsApp in Hindi, allowing them to route at-risk batteries to depots before they fail in customer hands.
How We Built It
ML Model: Random Forest classifier trained on 90 days of synthetic battery telemetry (voltage, temperature, current, charge cycles, internal resistance) Backend: Python Flask API with real-time WebSocket streaming Frontend: React dashboard with live battery grid and risk scoring Simulation: Generated realistic data for 100 batteries to demonstrate concept
Challenges We ran into
Creating realistic battery degradation patterns without access to Battery Smart's proprietary data Balancing prediction accuracy with false positive rate (optimized to 92% accuracy) Designing a simple interface for non-technical station partners.
Accomplishments that we're proud of
Built a working ML model that accurately predicts battery failures Created intuitive risk scoring (0-100) that anyone can understand Demonstrated real-time monitoring of 100 batteries simultaneously Calculated clear ROI: ₹3.6 Cr annual savings
What we learned
Battery degradation is complex: temperature, voltage, and charge cycles are all critical Real-time data streaming is essential for safety-critical applications Station partners need simple, actionable alerts—not raw data
What's next for SwapGuard AI
Integrate with Battery Smart's existing IoT infrastructure Deploy LSTM neural network for better long-term predictions Add blockchain ledger for battery lifecycle tracking Scale to 50,000+ batteries across 50 cities Implement automated warranty claims based on predicted failures
Built With
- flask
- iot
- machine-learning
- numpy
- pandas
- python
- react
- scikit-learn
- tailwindcss
- websocket
