AEE Collide

Real‑time arbitrage insights for data center power.

Comment

1. Project Title and Description

Dispatch IQ — Energy Dispatch Intelligence: An AI-powered decision tool for behind-the-meter (BTM) generators (e.g., data centers) that recommends IMPORT vs GENERATE every 15 minutes by comparing grid LMPs to on-site gas generation cost, with live spreads, forecasting, and operator-ready briefings.

2. Problem Statement

BTM operators face fast-changing power markets and fuel costs, but dispatch decisions are often manual, delayed, and uncertain. Wrong calls (especially during extreme events) can cause major cost overruns or missed profit opportunities, and it’s hard to translate volatile price signals into an actionable operating plan.

3. Solution Overview

Dispatch IQ computes a spread (LMP − generation cost) and turns it into:

  • Real-time recommendation (GENERATE/IMPORT)
  • 72-hour probabilistic forecast (Monte Carlo p10/p50/p90 bands)
  • Market regime detection (e.g., scarcity/heat dome/wind glut/winter storm)
  • AI operator briefing that summarizes what to do and why
  • Historical replay to validate behavior on real crisis scenarios

4. Working Prototype

  • GitHub repo: This project (FastAPI backend + React/Vite frontend)
  • Runs locally:
    • Backend: uvicorn backend.main:app on localhost:8000
    • Frontend: Vite dev server on localhost:5173
  • Demo endpoints/UI features: site map, spread ticker, forecast chart, schedule grid, briefing card, replay scenarios

5. Technical Implementation

  • Backend: Python + FastAPI, NumPy-based Monte Carlo forecasting, scenario/replay APIs, optional integrations (EIA gas prices, OpenAI for briefings)
  • Frontend: React + Vite, Recharts (forecast visualization), React-Leaflet (site selection map), terminal-style UI
  • Architecture: Frontend calls REST endpoints like /api/dispatch/current, /api/dispatch/forecast, /api/dispatch/briefing, /api/replay/{scenario}

6. Use Case Explanation (Innovation, Execution, Impact)

  • Innovation: Combines energy market signals + on-site generation economics + uncertainty (probabilistic forecast) + operator-grade AI summaries into a single dispatch workflow.
  • Execution: End-to-end working product—API + UI—covering live decisioning, forecasting, site comparison, and replayable validation.
  • Impact: Enables data centers/industrial sites to reduce energy spend and capture arbitrage, and make faster, defensible dispatch calls during volatility (e.g., Uri/heat dome/wind glut), improving both cost control and operational readiness.

Built With

Share this project:

Updates