11a55an / air-pollution-forecasting

Forecasting temperature and air pollutants like CO, NO2, SO2, O3, PM2.5, PM10 using different ML & DL models.

Envirocast: Realtime Air Monitoring and Forecasting

Envirocast is a real-time air monitoring and forecasting system that provides monitoring and forecasting of indoor and outdoor environmental parameters. The system includes an IoT component for data collection, a machine learning component for forecasting, and a mobile app developed with Flutter for user interface and alerts.

Features

• Real-time Monitoring:

  • Indoor Parameters: Temperature, Humidity, Dust, LPG, Natural Gas, Carbon Monoxide.
  • Outdoor Parameters: Temperature, Carbon Monoxide, Sulfur Dioxide, Nitrogen Dioxide, Ozone, PM2.5, PM10, AQI.

• Hourly Forecasting:

  • Outdoor parameters forecasted for the next 7 days.

• Alerts:

  • Notifications when monitored parameters reach harmful levels.

Components

IoT

• Hardware Used: Arduino Uno, MQ5, MQ7, MQ135, DHT11, GP2Y101AU0F sensors, ESP32 Wi-Fi Module.
• Functionality: Collects real-time data from sensors and sends it to Firebase.

Machine Learning

• Dataset: Gujrat, Pakistan dataset (Feb 2022 - June 2024).
• Models Trained: Moving Averages, ARIMA, SARIMA, FB Prophet, LSTM, and 1D-CNNs.
• Best Models: 1D-CNN models (8 models trained for 8 parameters).
• Deployment: Models deployed on AWS.

Mobile App

• Framework: Developed using Flutter.
• Features: Provides a user-friendly interface for monitoring, forecasting, and receiving alerts.

Repository Structure

• envirocast_IOT: Includes Arduino code for sensor interfacing.
• envirocast_ML: Contains notebooks/scripts for data preprocessing, model training, and evaluation.
• envirocast_App: Flutter code for the mobile application.

Screenshots

Contributors

• Meerab Irfan
• Hassan Tahir
• Sheraz Ahmed