Pravah provides electric utilities with a machine‑learning platform that delivers real‑time forecasts of demand and distributed generation, graph‑neural‑network models of grid topology, and computer‑vision mapping of assets from satellite and street‑level imagery. The system runs probabilistic simulations using reinforcement learning to identify overload risks and constraint failures before they occur, giving operators actionable alerts to improve reliability and reduce operational risk.
Funding
Funding not disclosed
Founders
Product
Problem
Utilities face increasing difficulty forecasting demand and generation due to extreme weather, electric vehicle adoption, and distributed rooftop solar, while their grid models rely on incomplete and noisy data, leading to risky operations and overloads.
Solution
Pravah applies machine learning to provide utilities with a real‑time, data‑driven view of grid behavior under stress. Its platform combines deep learning demand and generation forecasts, graph neural network‑based grid modeling, and computer‑vision asset mapping from satellite and street‑level imagery to fill blind spots in distribution networks. Probabilistic simulations powered by reinforcement learning generate thousands of possible grid futures, identifying overload risks and constraint failures before they occur. The integrated solution delivers actionable risk alerts and constraint insights, enabling operators to mitigate overloads and improve reliability without manual data collection.
Target Audience
Primary customers are electric utilities and grid operators responsible for distribution network planning, real‑time operations, and reliability management.
Features
- Deep learning models that forecast electricity demand and distributed generation across short‑term and long‑term horizons, capturing volatility missed by legacy methods
- Graph neural network (GNN) framework for accurate grid topology and constraint modeling using partial utility data
- Computer‑vision pipelines that extract grid asset locations and rooftop solar installations from satellite and street‑level images
- Reinforcement‑learning‑driven probabilistic simulations that evaluate thousands of grid scenarios to surface high‑risk overloads and failure modes
- Real‑time risk detection dashboards that highlight local constraints failing before system‑wide limits are reached
- Deployment experience with utilities in India, Germany, and the United States, demonstrating operational risk reduction in live systems