Real-time electricity demand forecasting for 30 Indian smart cities
Metrics
XGBoost + Linear Regression hybrid
9.12% MAPE on unseen test data
Lag-free feature pipeline using weather + calendar features, no historical demand inputs needed
FastAPI backend + Next.js dashboard
<300ms end-to-end prediction latency
Tech stack
Python
XGBoost
Scikit-learn
FastAPI
Next.js
Tailwind CSS
What it does
Vidyut AI forecasts electricity demand in real time across 30 Indian smart cities, using
a hybrid of XGBoost and Linear Regression that reaches 9.12% MAPE on unseen test data.
The feature pipeline is lag-free: it's built entirely from weather and calendar features,
with no historical demand inputs required. That means a forecast doesn't depend on having
an unbroken recent feed of past demand numbers to work from, only the weather forecast and
the calendar.
A FastAPI backend serves the model behind a Next.js dashboard, with end-to-end prediction
latency under 300ms. Both are live: the dashboard is deployed and the code is public on
GitHub.
Vidyut AI, forecasting electricity demand across 30 Indian smart cities
