Great Energy
Great Energy is an AI-native power trading company that systematically trades the gap between day-ahead and real-time electricity prices on the Texas ERCOT grid. Its fully automated pipeline ranks approximately 24,000 node-hours daily using 175 signals per node, including AI weather ensembles, load forecasts, and congestion shadow prices, with zero manual discretion from data ingestion to settlement.
- Artificial Intelligence
- Energy
Funding
Founders
Founder details are not available yet.
Product
Problem
Wholesale electricity markets price power twice—day-ahead and real-time—and the gap between these prices creates systematic trading opportunities. However, the complexity of market data, weather uncertainty, and the sheer number of nodes make this largely inaccessible to manual or rules-based trading approaches.
Solution
Great Energy is a pure systematic power trading company operating on the Texas ERCOT grid. Every morning, its AI ranks approximately 24,000 node-hours by the probability that day-ahead and real-time prices will diverge, then trades the most confident positions by selling day-ahead and buying back in real time. The system pulls 175 signals per node—including AI weather ensembles, ERCOT load and renewable forecasts, outage data, and congestion shadow prices—and runs fully automated from data ingestion to settlement with zero manual discretion. The approach was validated by testing 4.2 billion combinations against a null model built to fool the same procedure.
Target Audience
Primary customers are wholesale electricity market participants, including proprietary trading desks and financial institutions, seeking systematic exposure to day-ahead vs. real-time price spreads in ERCOT and other US grids.
Features
- AI-powered ranking of ~24,000 node-hours daily across ERCOT, scoring each by probability of day-ahead vs. real-time price divergence
- 175 signals per node-hour, including multiple weather forecast families, ERCOT load and renewable forecasts, congestion shadow prices, outages, and system lambda
- Fully automated pipeline from data ingestion to settlement, with all model inputs stored exactly as they stood before the trading deadline to prevent look-ahead bias
- Rigorous validation framework: 4.2 billion combinations tested against a null distribution, walk-forward testing, and strict cutoff enforcement
- Automated guardrails including loss halts, cold-weather stand-downs, worst-day size limits, and a human stop switch
- Market expansion roadmap to PJM, MISO, SPP, CAISO, NYISO, and New England, with each market earned on its own data before capital follows