Prompt-to-Protocol

From Prompt to Protocol: Fast Charging Batteries with Large Language Models

Environment Setup

Prerequisites

• Python 3.9+
• CUDA (optional, for GPU acceleration)

Installation

1. Clone the repository:

git clone https://github.com/your-repo/prompt-to-protocol.git
cd prompt-to-protocol

2. Create a virtual environment:

conda create -n charging python=3.12.3
conda activate charging

3. Install dependencies:

For Case 1 & Case 2 (Constant / Predefined Heating Optimization):

pip install -r requirements.txt

Note: Case 1 & Case 2 require pybamm==24.1

For Case 3 (Adaptive Charging Optimization):

pip install ax-platform==0.4.0 deap==1.4.1 matplotlib==3.8.2 "numpy<2" openai==1.51.2 optuna==4.0.0 pandas==2.2.3 scipy==1.14.1 termcolor==2.5.0
pip install pybamm==25.1.1

Note: Case 3 requires pybamm==25.1.1 due to API changes in PyBaMM

API Key Configuration

For LLM-based optimization, you need to configure your OpenAI API key:

Option 1: Using a .env file (Recommended)

1. Create a .env file in the project root:

touch .env

2. Add your API key to the .env file:

export OPENAI_API_KEY="your-api-key-here"

3. Load the environment variables before running scripts:

source .env

Quick Start: Running MWE (Minimum Working Example)

The scripts/ directory contains self-contained examples that you can run directly:

MWE Simulation

Test the battery simulation:

python scripts/MWE_simulation.py

MWE Optimization

Run a complete SAASBO optimization:

python scripts/MWE_optimization.py

This will:

• Initialize a neural network for charging current control
• Run 30 Sobol initialization trials
• Run 60 batches of SAASBO optimization (3 samples per batch)
• Save results to ./experiments/MWE_results/

Project Structure

prompt-to-protocol/
├── scripts/                    # MWE scripts (standalone examples)
│   ├── MWE_optimization.py     # Complete optimization example
│   └── MWE_simulation.py       # Simulation example
├── src/
│   ├── p2o/                    # Prompt-to-Optimizer
│   │   ├── evaluate_model/     # Optimization methods
│   │   │   ├── SAABO_constant_heating.py   # Case 2
│   │   │   ├── SAABO_adaptive_RNN.py       # Case 3
│   │   │   ├── ECM_gradient_descent.py      # Case 1 
│   │   │   └── random_constant_heating.py   # Case 2
│   │   ├── simulation/         # Battery simulation
│   │   │   ├── sim_p2o_c1_c2.py  # Case 1 & 2 simulation
│   │   │   └── sim_p2o_c3.py     # Case 3 simulation
│   │   ├── evolution/          # Evolutionary algorithms
│   │   ├── evaluation/         # Evaluation scripts
│   │   └── llm_generation/     # LLM-based generation
│   ├── p2p/                    # Prompt-to-Protocol
│   │   ├── simulation/
│   │   ├── evolution/
│   │   └── llm_generation/
│   └── tools/                  # Utility functions
├── experiments/                # Experiment results
├── plots/                      # Visualization notebooks
├── requirements.txt
└── README.md

License

See LICENSE for details.