Skip to content

AlphaSearch: Agentic Energy Arbitrage with Battery Storage ⚡

AlphaSearch is an agentic AI pipeline for daily battery arbitrage in wholesale electricity markets.

It combines:

  • A MILP oracle (via agentic_energy) for exact charge/discharge schedules
  • RL and heuristic controllers for adaptive operation
  • CrewAI agents + MCP tools for orchestration, reasoning, and visualization
  • A Streamlit UI for interactive exploration and “chat with your battery” style workflows

Core Idea

You bring:

  • Historical or live price data
  • A storage asset (capacity, power limits, efficiencies)

AlphaSearch:

  1. Forecasts prices for the next day
  2. Chooses an optimization agent (MILP, heuristics, RL, or LLM-based)
  3. Generates a schedule of when to charge, discharge, or stay idle
  4. Explains decisions with plots and natural-language reasoning

You interact with everything through a simple chat box and a results panel.

Who Is This For?

  • Energy & power systems researchers
  • Quant / trading / storage optimization teams
  • People interested in agentic AI applied to real physical systems

If you want to understand the repo structure and launch the app, start with the Quickstart.