EnerVision - AI Energy Forecasting for Indian Buildings

AI-powered energy forecasting system using time series foundation models to predict consumption, detect anomalies, and reduce costs by 20% for Indian commercial buildings.

Comment

⚡ EnerVision – AI Energy Forecasting for Indian Buildings About the Project

EnerVision is an AI-driven prototype for building energy forecasting, anomaly detection, and efficiency recommendations.

Developed during the IndiaAI Impact Gen-AI Hackathon, this project was inspired by India’s urgent need for intelligent energy management tools. With rising electricity demand, rapid air conditioning adoption, and the drive toward sustainability, we envisioned a platform that leverages AI + time-series modeling to help users cut costs, reduce carbon footprint, and support India’s carbon neutrality goals.

🌱 Inspiration

We were motivated by a pressing challenge: most Indian buildings lack intelligent energy monitoring systems. As renewable energy integration grows, short-term forecasting and anomaly detection become crucial.

👉 The hackathon gave us the perfect stage to explore how Generative AI + time-series models could create a smart, scalable solution.

⚡ What It Does (Prototype Scope)

EnerVision’s prototype focuses on four key features:

Load Forecasting

Predicts short-term (24h–7d) electricity demand.

Accounts for seasonal/holiday effects and peak hours.

Anomaly Detection

Flags unusual consumption (e.g., faulty AC or lighting).

Efficiency Advisor

Suggests demand shifting to off-peak hours.

Recommends solar/battery usage for sustainability.

Interactive Dashboard (Concept)

Visualizes forecasts & anomalies in simple charts.

Prototype included mock UI with planned voice assistant support.

🛠️ How We Built It (in 3 Days 🚀)

Day 1:

Brainstormed concept & finalized scope.

Set up baseline forecasting using Prophet with sample datasets.

Day 2:

Experimented with TSFMs (time-series foundation models).

Implemented a simple anomaly detection pipeline.

Day 3:

Built a prototype dashboard using React.

Integrated APIs and drafted Advisor Agent logic.

Prepared demo presentation + submission.

🚧 Challenges We Faced

Time crunch ⏳: Building an AI-driven prototype in 3 days meant making trade-offs.

Data scarcity: Real building data was limited, so we relied on sample datasets + open data.

Integration: Full agent orchestration was conceptual, not fully deployed.

UI/UX: With limited time, we focused on mock dashboards rather than a polished product.

🏆 Accomplishments We’re Proud Of

Designed a clear, scalable architecture (Forecasting Agent + Anomaly Agent + Advisor Agent).

Delivered a working prototype with forecasting + anomaly detection in just 3 days.

Created a user-friendly concept dashboard to make energy data accessible.

Positioned EnerVision as a solution aligned with India’s clean energy goals.

📚 What We Learned

TSFMs are game-changing for time-series problems.

Building multi-agent AI workflows can make insights more actionable.

Learned to balance ambition with feasibility in hackathon timelines.

Gained hands-on experience in rapid prototyping under pressure.

🚀 What’s Next for EnerVision

Integrating with IoT smart meters for real-time data.

Running demand response simulations with renewable integration.

Adding benchmarking to compare across buildings.

Scaling from prototype → pilot projects with housing societies & commercial complexes.

🏗️ Built With

Languages: Python, JavaScript

Frameworks: PyTorch, FastAPI, React

Tools: Prophet, Isolation Forest, Autoencoders, TSFMs

Platforms: IBM watsonx.ai (planned), Docker (concept)

Databases: PostgreSQL, Redis (planned)

🔗 Try It Out

Demo Site (Prototype): http://localhost:3000/

GitHub Repo: https://github.com/cookiethecat-psst/enervision-ibm-granite-forecasting

Video Demo: [Link to your demo video]

Built With

Share this project:

Updates