Adaptive Control in Smart Microgrids using Deep Reinforcement Learning
This project demonstrates the application of reinforcement learning (RL) to control a simplified smart microgrid environment. Inspired by real-world challenges in industrial automation and energy efficiency, the agent learns how to balance 01) energy demand, 02) battery storage, and 03) renewable generation through sequential decision-making.
--> simulates a microgrid environment with changing energy inputs (solar, wind, battery) and usage demands. The RL agent learns a policy to balance storage, consumption, and external supply, optimising for energy efficiency and cost over time.
this makes use of proximal policy optimisation to:
- reduce reliance on external grid power
- minimise operational costs
- maintain energy balance + battery health
- smart_microgrid_env.py = custom OpenAI gym-compatible env't for microgrid
- ppo_agent.py = proixmal policy optimisation agent w/ PyTorch
- train.py = training script for RL agent
- requirements.txt = dependencies for project
install dependencies:
pip install -r requirements.txt