GridSense AI: Adaptive EV Charging Intelligence Layer

GridSense AI is an advisor-layer prototype for BESCOM that predicts demand, nudges charging behavior, plans infrastructure, and explains every recommendation. It does not modify the grid or control chargers. It uses synthetic data to remain safe and non-intrusive.

What it does

• Predicts EV charging demand by zone and hour.
• Detects peak periods and sends peak alerts.
• Performs soft demand shifting with incentive-aware nudges.
• Plans infrastructure using a future demand score and grid headroom.
• Produces explainable, actionable outputs in CSV and JSON.

Closed-loop flow

Predict -> Optimize -> Recommend -> Explain -> Improve

How to run

1. Install dependencies:

pip install -r requirements.txt

2. Run the pipeline:

python -m src.main

Dashboard

Run the Streamlit UI to explore outputs and trigger pipeline runs:

streamlit run dashboard/app.py

The dashboard includes filters, a station siting map, pipeline controls, and PDF/PNG export snapshots.

Outputs

The run creates an outputs/ folder with:

• demand_forecast.csv
• peak_periods.csv
• peak_explanations.csv
• charging_recommendations.csv
• zone_summary.csv
• zone_hotspots.csv
• station_siting.csv
• feature_importance.csv
• summary.json

Notes

• All data is synthetic and masked by design.
• The prototype is a decision-support layer only.
• Explainability is provided via feature importance and rule-based explanations.