AI Data Center Microgrid with Battery Storage Peak-Shaving Controller

📌 Project Overview

Artificial Intelligence data centers exhibit highly dynamic, high-power consumption profiles—especially during massive parallel LLM training workloads. This project designs and simulates a Rule-Based Logic Controller that dispatches a Battery Energy Storage System (BESS) to shave peak power demands, ensuring grid/substation stability and reducing peak demand utility charges.

This entire framework is built in Python, designed to run completely open-source and cloud-hosted via Google Colab to maintain a zero-local-storage environment footprint.

📊 Key Features

• Dynamic Load Profiling: Simulates real-world AI data center loads, combining steady-state cooling/compute overhead with stochastic training spikes.
• Smart Throttling/Dispatch Logic: Actively flattens substation power demand when load exceeds a predefined safety threshold ($20\text{ MW}$).
• BESS Constraints Management: Dynamically models State of Charge (SoC) bounds ($20% - 100%$), charging efficiencies, and maximum charge/discharge rates.

📂 Repository Structure

├── notebooks/
│   └── peak_shaving_simulation.ipynb   # Main Google Colab Notebook
├── simulation_results.png              # Output visualization graph
└── README.md                           # Documentation Portfolio