SmartSense

Plug-and-play AI for energy forecasting and anomaly detection, powering smarter, greener cities.

Comment

Inspiration

We noticed that while large corporations have access to powerful forecasting tools, public institutions like schools, hospitals, and municipalities often lack affordable solutions. With India’s Net-Zero 2070 goal in mind, we built an open-source system that makes energy forecasting accessible, scalable, and free for everyone.

What it does

SmartSense provides short-term energy load forecasting (hourly/daily) and real-time anomaly detection. It helps institutions detect abnormal energy spikes, manage consumption efficiently, and integrate renewables—all without expensive infrastructure.

How we built it

  • Used Darts and Nixtla NeuralForecast for lightweight forecasting models.
  • Pulled free weather data from OpenWeatherMap to improve prediction accuracy.
  • Designed a modular, TSFM-ready architecture so that future advanced models can be plugged in seamlessly.
  • Deployed the system on HuggingFace Spaces, Vercel, and Render for zero-cost, easy adoption.

In simple terms, SmartSense looks at past energy usage patterns along with weather conditions (like temperature and humidity) to predict upcoming energy demand and flag unusual spikes.

Challenges we ran into

  • Limited availability of public energy datasets.
  • Balancing high accuracy with the need to run on low-compute, CPU-only setups.
  • Integrating forecasting, weather data, and anomaly detection into one reliable pipeline within the hackathon time frame.

Accomplishments that we're proud of

  • Achieved 20–30% lower forecast errors compared to traditional baselines like Prophet/ARIMA.
  • Built a tool that runs smoothly on free cloud tiers and even basic laptops.
  • Designed a plug-and-play system that public institutions can adopt at zero cost.

What we learned

  • Data quality and preprocessing can be more important than complex models.
  • Accessibility > Complexity when designing for real-world impact.
  • A scalable, modular design ensures the project can evolve even after the hackathon.

What's next for SmartSense

  • Integrate next-gen Time Series Foundation Models (Chronos, TimesFM).
  • Extend to renewable energy optimization (solar, wind).
  • Build a city-wide dashboard for municipalities to monitor and optimize energy usage in real time.
  • Explore partnerships with government bodies under the Smart Cities Mission.

Built With

Share this project:

Updates