Energy Simulation for Battery Optimization

This project simulates how a compute cluster could intelligently decide when to charge a battery, run on grid power, or run on battery using real grid data.

The system pulls real hourly electricity data from the U.S. Energy Information Administration (EIA) and combines it with simulated compute demand and battery behavior to model decision-making for energy-aware infrastructure.

This prototype is part of an exploration into energy-aware scheduling for AI / GPU workloads.

---

Project Goal

The goal is to develop a model that:

1. Uses real grid signals
2. Simulates compute infrastructure power demand
3. Simulates battery constraints
4. Makes hourly decisions on energy usage

The decision engine determines whether the system should:

• Charge the battery
• Run compute from the grid
• Run compute from the battery

---

Data Sources

Real Data

Source: EIA API
https://www.eia.gov/opendata/

Currently pulled:

• Hourly grid demand (used as a proxy for electricity price)

Future integration:

• Real hourly electricity prices
• Demand forecasts

---

Simulated System Components

Battery Model

Variable	Description
Battery Charge	0–100% battery state
Charge Rate	kW added per hour
Discharge Rate	kW consumed per hour
Min Threshold	Battery never below 20%
Max Threshold	Battery never above 95%

Compute Demand Model

Simulated GPU cluster metrics:

Variable	Range
GPU Utilization	0–100%
Jobs in Queue	0–50 jobs
Power Draw	Derived from utilization

---

Decision Engine

Every hour the model evaluates:

• Grid demand (price proxy)
• Compute demand
• Battery state

Possible decisions:

State	Condition
Charge battery	Grid energy is cheap and battery is not full
Run off grid	Grid energy is cheap and compute demand is high
Run off battery	Grid energy is expensive and battery has charge

---

Visualization

The simulation outputs a 30-day visualization showing:

• Grid demand signal
• Battery charge level
• Decision behavior over time

Example output:

---

Project Structure

energy-sim-amply/
├── simulation.py          # Main simulation script
├── README.md              # Project documentation
└── requirements.txt       # Python dependencies

---

Installation

Clone the repository:

git clone https://github.com/moukthika-gunapaneedu/energy-sim-amply.git
cd energy-sim-amply

Install dependencies:

pip install pandas numpy matplotlib requests

---

Environment Setup

Set your EIA API key as an environment variable:

export EIA_API_KEY=your_api_key_here

You can obtain a free key here: https://www.eia.gov/opendata/

---

Run the Simulation

python simulation.py

The script will:

1. Pull 30 days of hourly data
2. Simulate compute + battery behavior
3. Run the decision engine
4. Generate a visualization

---

Current Status

Completed:

• EIA API integration
• Battery simulation
• Compute demand simulation
• Hourly decision engine
• 30-day visualization

---

Next Steps

Planned improvements:

• Integrate real hourly electricity price data
• Add grid demand forecasting
• Incorporate time-of-day and weekday patterns
• Improve decision logic with optimization instead of threshold rules
• Evaluate cost savings vs naive scheduling

---

Technologies Used

• Python
• Pandas
• NumPy
• Matplotlib
• Requests
• EIA Energy API

---

Team: Amply