Home

RL-ADN

RL-ADN is a Python library for reinforcement learning research on energy storage dispatch in active distribution networks. It combines environment simulation, packaged feeder data, baseline optimization components, and the Laurent power-flow solver in one research-friendly codebase.

Features

• Versatile Benchmarking: Model diverse energy arbitrage tasks with full flexibility.
• Laurent Power Flow: Over 10 times faster computational speed compared to traditional methods.
• Seamless Transition: Designed for both simulated environments and real-world applications.
• Open-source: Easily accessible for modifications, customizations, and further research.
• Phase A Topology Scenarios: 34-bus and 69-bus feeders now support hand-authored radial topology scenarios that switch on reset().

Phase A Topology-as-Scenario

RL-ADN now supports topology variation as an environment scenario, aligned with the TP1-TP7 evaluation style used in the topology-aware GNN transferability paper. In this phase:

• topology is not part of the RL action space
• topology changes only on reset()
• 34-bus and 69-bus feeder scenario libraries are provided
• graph and topology metadata can be exported for later GNN work

Installation

To install RL-ADN:

pip install RL-ADN

Or install from git:

git clone https://github.com/EnergyQuantResearch/RL-ADN.git
cd RL-ADN
pip install -e .

Quick Start

from rl_adn import PowerNetEnv, make_env_config

config = make_env_config(node=34, topology_scenario="TP2", return_graph=True)
env = PowerNetEnv(config)
state, info = env.reset(return_info=True)
print(info["topology_scenario"])

To sample from multiple topologies:

config = make_env_config(
    node=34,
    topology_mode="scenario_pool",
    topology_pool=["TP2", "TP3", "TP4"],
    return_graph=True,
)

Where To Go Next

• Getting Started
• Environment Configuration
• Topology Scenarios
• Graph Interface
• Tutorials

Contributing

Potential directions include richer ESS models, topology-aware learning, and broader benchmark coverage across feeders and operating conditions.

License

MIT License