FleetRL - Realistic RL environments for commercial EV fleets

Github: https://github.com/EnzoCording/FleetRL

Software features:

• Base-derived class architecture and easily interchangeable configurations

• Modular implementation approach for easily extendable model and methods

• PEP8 compliant (unified code style)

• Documented functions and classes

• Stable-Baselines3 integration, parallelization, tensorboard support

• Extensive data logging and evaluation possibilities

Unique Implementations:

• Non-linear battery degradation

• Fleet schedule generation - inspired by emobpy and applied to commercial fleets

• Bi-directional charging

• Building load, grid connection limit, PV, spot price

• Arbitrarily variable episode length, 15-min time resolution

• Benchmarking of RL methods with static benchmarks: uncontrolled charging, night charging, distributed charging

• Benchmarking of RL methods with linear optimization

User Guide

• Overview
• Installation
• Creating a custom environment
  • Time-series
  • Static inputs
  • Summary
• Agent Training
• Agent Evaluation
• Benchmarking

Modules

• Fleet environment
• Benchmarking
• Agent evaluation
• Data logger
• Load calculation
• Normalisation
• Observation
• Schedule
• Time picker
• Data processing
• EV charging
• Battery degradation
• Event manager
• Rendering

Indices and tables

• Index

• Module Index

• Search Page