Urja-Drishti

Urja-Drishti: Hyper-Contextual Energy Load Forecasting and Management in Indian Buildings

Abstract: From Data Points to Intelligent Decisions

Accurate short-term energy load forecasting is crucial for decarbonizing India's urban building sector, but generic models fail to account for the nation's diverse climate, construction, and grid conditions. Project Urja-Drishti offers an innovative agentic AI framework using a pre-trained Time Series Foundation Model (TSFM).

Our system uses an intelligent agent. We fine-tune a base TSFM on datasets representing distinct Indian building archetypes (e.g., Bengaluru offices, Delhi residences, Chennai hospitals) for hyper-contextual accuracy. This fine-tuned model becomes the "brain" for an AI agent, built with IBM's Agent Development Kit (ADK), performing three critical tasks:

Delivering highly accurate, zero-shot energy load forecasts for new buildings within a known archetype.

Detecting subtle anomalies indicating equipment malfunction or energy wastage.

Generating actionable, natural-language recommendations for building managers.

Urja-Drishti provides intelligence, empowering facility managers to optimize energy, reduce costs, and contribute to India's decarbonization goals.