AI Solar Tracking System

Machine-learning assisted solar tracking platform featuring simulation engineering, FastAPI deployment, and an interactive web dashboard.

This project combines mathematical modeling, simulation engineering, and machine learning to optimize solar panel orientation based on the Sun direction vector.

This project demonstrates how mathematical modeling, machine learning, backend engineering, and simulation systems can be integrated into a single real-time engineering workflow.

---

Highlights

• Stewart-platform inspired solar tracking simulation
• Six-leg parallel mechanism architecture
• Machine learning angle prediction using Random Forest Regressor
• R² Score > 0.99
• Average prediction error ≈ 0.1°
• FastAPI REST API deployment
• Interactive frontend dashboard
• Constraint-aware tracking logic
• Mechanical tilt and actuator limit validation

---

Live Demo

Frontend: https://sun-tracking-six-leg-panel-system.vercel.app

Backend API: https://sun-tracking-six-leg-panel-system.onrender.com

API Docs: https://sun-tracking-six-leg-panel-system.onrender.com/docs

Demo Preview

Web Interface

Simulation

ML Accuracy

---

Project Architecture

Sun Vector
    ↓
ML Model (Random Forest)
    ↓
Predicted Zenith
    ↓
Constraint Validation
    ↓
3D Simulation Engine
    ↓
Frontend Dashboard

---

Machine Learning Model

Input Features

• sun_x
• sun_y
• sun_z

Target

• required_zenith_deg

Model

• Random Forest Regressor

Performance

• R² > 0.99
• Average prediction error ≈ 0.1°

Engineering Improvement

Resolved real-world inference mismatch by aligning training data normalization with production API inputs.

---

Challenges & Lessons Learned

• Handling physically impossible tracking states caused by actuator constraints
• Preventing unstable panel motion near maximum tilt angles
• Aligning ML training data normalization with production API inputs
• Designing a modular architecture between simulation, ML inference, and backend API
• Improving prediction stability during real-time simulation updates

---

Simulation Features

• Real-time animated Sun movement
• Dynamic panel orientation tracking
• Six-leg actuator geometry
• Zenith tilt constraints
• Leg length limit detection
• Color-coded system states
• Live diagnostics panel

---

Tech Stack

Languages & Libraries

• Python
• NumPy
• Pandas
• Matplotlib
• scikit-learn

Backend

• FastAPI
• Uvicorn
• SQLite

Frontend

• HTML
• CSS
• JavaScript

Tools

• Git
• GitHub
• PyCharm

Deployment

• Render
• Vercel
• Docker

---

API Usage

Run Locally

uvicorn src.api:app --reload

Endpoints

POST /predict

Predicts the required panel zenith angle from the given Sun vector.

GET /history

Returns recent prediction history including:

• predicted zenith
• analytic zenith
• prediction error

Example Request

{
  "sun_x": 35,
  "sun_y": 69,
  "sun_z": 42
}

Example Response

{
"predicted_zenith": 61.32,
"analytic_zenith": 61.50
}

---

How to Run

Clone Repository

git clone https://github.com/aleynaozdogann/sun-tracking-six-leg-panel-system.git cd sun-tracking-six-leg-panel-system

Install Dependencies

pip install -r requirements.txt

Train Model

python src/train_model.py

Run API

uvicorn src.api:app --reload

Run Simulation

python src/main.py

---

Why This Project Matters

This project demonstrates the combination of:

• Machine learning in production
• Real-time simulation systems
• Mathematical modeling
• Backend API engineering
• Frontend integration
• Debugging real-world ML deployment issues

---

Next Engineering Goals

• Reinforcement learning based adaptive tracking
• Real-time monitoring dashboard
• Hardware actuator integration
• GPU accelerated simulation
• Cloud-native deployment pipeline

---

Author

Aleyna Özdoğan

Mathematics graduate focused on:

• Artificial Intelligence
• Backend Development
• Simulation Engineering
• Computational Modeling