An innovative self-adjusting solar panel system that maximizes energy absorption using computer vision and deep learning.
This repository contains the full implementation of our AI-powered solar panel tracking system. The project optimizes solar energy capture by accurately tracking the sun's position in real-time, using a lightweight deep learning model running on a Raspberry Pi with motor control via microcontroller.
The system employs a sophisticated computer vision pipeline built with OpenCV and Ultralytics YOLOv8:
- Image Acquisition: Captures frames from a camera module at dynamically calculated intervals
- Sun Detection: Processes frames through our custom-trained YOLOv8 model
- Position Analysis: Calculates the sun's offset from center using pixel coordinates
- Motor Control Commands: Translates positional data into servo motor instructions
def calculate_distance(center_x, center_y, bbox):
"""Calculates the distance of the detected object from the center."""
x1, y1, x2, y2 = bbox
object_center_x = (x1 + x2) / 2
object_center_y = (y1 + y2) / 2
distance_x = object_center_x - center_x
distance_y = object_center_y - center_y
return distance_x, distance_y
The system intelligently manages detection frequency based on weather conditions:
- Clear Sky: 60-second intervals for optimal tracking
- Partly Cloudy: 180-second intervals to balance accuracy and power usage
- Overcast/Rainy: 300-second intervals to conserve power during low solar output
- Nighttime: Extended sleep mode until sunrise to maximize energy efficiency
def calculate_next_interval():
"""Calculate the next interval time based on weather conditions and time of day"""
# Check if it's nighttime (after sunset or before sunrise)
is_nighttime = current_time > sunset or current_time < sunrise
if is_nighttime:
# If sun is set, use a very long interval until next sunrise
time_until_sunrise = sunrise - current_time if current_time < sunrise else (sunrise + 86400) - current_time
new_interval = min(int(time_until_sunrise), 3600) # Cap at 1 hour max
else:
# Base interval on weather conditions during daytime
if "clear" in weather_condition or cloud_coverage < 20:
new_interval = 60 # 1 minute
elif "cloud" in weather_condition or cloud_coverage < 70:
new_interval = 180 # 3 minutes
else:
new_interval = 300 # 5 minutes- Real-time Sun Position Detection: Utilizes a custom-trained YOLOv8 model to detect the sun's position with 98.8% mAP50-95 accuracy
- Intelligent Weather Adaptation: Dynamically adjusts detection intervals based on cloud coverage and time of day
- Day/Night Cycle Awareness: Conserves power by entering low-power mode during nighttime hours
- Comprehensive Monitoring Dashboard: Next.js web interface for real-time system monitoring and control
- Robust Data Logging: Firebase integration for performance tracking and system diagnostics
- Automated Fallback Mechanisms: Implements non-ML algorithms for tracking during adverse weather conditions
- Optimized Power Management: Camera activates only when needed to conserve energy
The Next.js dashboard provides comprehensive system monitoring and control:
- System Status Overview: Camera state, interval settings, and weather conditions
- Sun Position Visualization: Real-time graphical representation of detected sun position
- Performance Metrics: CPU, memory, and disk usage monitoring
- Weather Data Display: Current conditions with sunrise/sunset visualization
- Control Panel: Camera activation, interval adjustment, and test mode controls
The system uses Firebase Firestore for comprehensive data logging:
- ModelLog Collection: Stores detection results, system performance metrics, and timestamps
- ProgramLog Collection: Records weather data, interval calculations, and system events
- Test Mode Data: Captures continuous testing data for system optimization
- Flask: Powers the RESTful API server with endpoints for system control and monitoring
- OpenCV: Handles image processing and camera operations with efficient frame manipulation
- Ultralytics YOLOv8: Provides the core object detection capabilities with state-of-the-art accuracy
- Supervision: Simplifies detection visualization and post-processing
- Firebase Admin SDK: Enables secure database operations for logging and monitoring
- Flask-CORS: Ensures secure cross-origin resource sharing for the web dashboard
- python-dotenv: Manages environment variables for secure API key storage
- Next.js: Creates a responsive, server-rendered React application
- TailwindCSS: Provides utility-first CSS for rapid UI development
- NextUI: Delivers modern UI components with accessibility features
- Firebase SDK: Enables real-time data synchronization with the backend
Our custom-trained YOLOv8 model achieves exceptional performance for sun detection:
| Model | mAP50-95 | Precision | Recall | Images | Type | Suitability for Microcontroller |
|---|---|---|---|---|---|---|
| Our YOLO model v3 | 98.8% | 91.7% | 92.4% | 3409 | YOLOv8 | ✅ Excellent (Lightweight) |
| sun-tracking-555mn/4 | 97.7% | 93.0% | 89.2% | 923 | YOLOv8n | ✅ Good (Roboflow Library) |
| solar-re1fe/1 | 96.8% | 95.2% | 94.3% | 2684 | Roboflow 3.0 | ✅ Good (Optimized for Edge) |
| sun-tracking-photovoltaic-panel/1 | 98.2% | 93.7% | 93.7% | 196 | Roboflow 2.0 | |
| sun-tracking/3 | 92.5% | 94.7% | 91.8% | 1090 | YOLOv5 | ✅ Compatible (Proven on MCUs) |
| Our YOLO model v2 | 42.6% | 78.7% | 79.3% | 274 | YOLOv8 | ❌ Too Heavy |
| Our YOLO model v1 | 21.3% | 58.0% | 64.0% | 198 | YOLOv8 | ❌ Not Viable |
The model was trained on a comprehensive dataset combining:
- Sun Tracking Photovoltaic Panel Dataset
- Sun Dataset
- Sun Tracking Dataset
- Solar Dataset
- Sun Detection Dataset
- Custom dataset with 400 manually labeled images
The Flask server exposes several RESTful endpoints:
- PUT /start_stop_camera: Toggles camera and model detection on/off
- PUT /change_interval: Updates the detection interval timing
- POST /test_model: Activates continuous testing mode for system validation
- GET /status: Returns comprehensive system status information
-
Clone the repository
git clone https://github.com/ashworks1706/AI-Solar-Panel.git cd AI-Solar-Panel -
Create a virtual environment with Python 3.11
cd python python3.11 -m venv venv source venv/bin/activate # On Windows: venv\Scripts\activate
-
Install dependencies
pip install -r python/requirements.txt
-
Configure environment variables for weather API
- Create a
.envfile in the project root with your API keys:
WEATHER_API_KEY=your_openweather_api_key - Create a
-
Set up Firebase
- Get firebase-secret.json from your Firebase project
- Place it in your root directory
-
Start the Flask server
python main.py
-
Run the dashboard (in a separate terminal)
cd .. npm install npm run dev -
Configure hardware components
- Connect Raspberry Pi to camera module
- Set up the microcontroller for motor control
- Ensure proper power connections for all components
The dashboard will be available at http://localhost:3000, and the Flask API will be running at http://localhost:5000.
The system includes a dedicated test mode that:
- Continuously captures and processes frames
- Logs detection results to Firebase
- Monitors system performance metrics
- Validates weather data integration
- Tests interval calculation algorithms
This comprehensive testing approach ensures reliable operation in various conditions.
- Integration with machine learning for predictive weather analysis
- Power consumption optimization through advanced sleep modes
- Enhanced motor control algorithms for smoother tracking
- Mobile application for remote monitoring and control
- Integration with smart home systems via MQTT
- @Ampers8nd (Justin Erd.): Mechanical design
- @Zhoujjh3 (Justin Zhou): Electrical components
- @ashworks1706 (Ash S.): Deep learning development
This project is licensed under the MIT License - see the LICENSE file for details.