Methanol APC Environment

A production-grade digital twin of an ICI Low-Pressure methanol synthesis reactor for reinforcement learning.

Live Demo on HuggingFace GitHub Repository

What is this?

An OpenEnv-compatible RL environment where an AI agent acts as an autonomous Advanced Process Control (APC) operator for a methanol synthesis reactor. The agent controls 13 plant variables across 5 stages to maximize profit while preventing thermal runaway and catalyst degradation.

Key Features

Feature	Details
Physics	5 kinetic models (LHHW, Graaf, VBF, Seyfert, Nestler), RK4 ODE, SRK EOS
Tasks	12 scenarios from Easy to Expert
Multi-Agent	4 agent classes (Reformer, Synthesis, Purification, Supervisory)
MCP Tools	Energy pricing, catalyst status, maintenance, emissions
Training	TRL + Unsloth GRPO bridge, Gymnasium wrapper
Integrations	DWSIM, Cantera, ChemSep, Azure Digital Twins
Deployment	Docker, K8s, HuggingFace Spaces, CI/CD
Tests	86 tests, 92% coverage

Quick Start

from methanol_apc_env import MethanolAPCEnv, MethanolAPCAction

async with MethanolAPCEnv.from_env("glitchfilter/methanol-apc-env").connect() as env:
    obs = await env.reset(task_name="optimization")
    action = MethanolAPCAction(feed_rate_h2=5.0, feed_rate_co=2.5,
                                cooling_water_flow=40.0, compressor_power=65.0)
    obs = await env.step(action)