neural-pid-controller

🔧🎯🔁🧠 Research on advanced process control: classical PID with enhancements vs a neural controller trained with backpropagation.

📌 Features

• Classical PID with:
  • Setpoint weighting (β, γ)
  • Anti-windup
  • Derivative filtering
  • Output limits & rate limits
  • Bumpless transfer
• Up to 3 feedforward signals
• Support for up to 5 disturbances
• Neural PID (NPID):
  • Backpropagation-inspired controller
  • Gradient descent optimization of control performance
  • Learns contribution of feedforwards and components
• Plant models:
  • FOPDT
  • Flash / column dynamics

🚀 Run

python -m venv .venv && . .venv/Scripts/activate
pip install -r requirements.txt
python experiments/compare.py