🌫️ Real-Time Air Quality Monitoring System (AQI-IoT-ML)

This project is an IoT + Machine Learning powered air quality monitoring system built using:

• ESP32
• MQ135 Gas Sensor
• DHT11 Temperature & Humidity Sensor
• Python ML pipeline
• Real-time data logging & AQI category prediction

It uses a combination of historical air quality datasets + live sensor data to build a model that predicts AQI categories, such as:

• Good
• Moderate
• Unhealthy
• Very Unhealthy
• Dangerous
• Hazardous

Live readings from sensors are processed through the trained model and the output is displayed + logged.

---

📁 Project Structure

aqi-iot-ml/
│
├── data/
│   ├── raw/
│   │   └── final_dataset.csv                  # Original dataset
│   └── processed/
│       ├── clean_normalized_dataset.csv       # After preprocessing
│       ├── scaler.pkl                         # MinMaxScaler for input normalization
│       ├── aqi_model.pkl                      # Trained ML model
│       └── label_encoder.pkl                  # Converts labels ↔ numbers
│
├── data/live/
│   └── my_live_dataset.csv                    # Your own real sensor data (auto-generated)
│
├── ml/
│   └── train_model.py                         # Trains the ML model
│
├── arduino/
│   └── esp32_mq135_dht/
│       └── esp32_mq135_dht.ino                # ESP32 sensor code
│
├── preprocess.py                               # Cleans + normalizes dataset
├── sensor_predict.py                           # Reads ESP32 serial → predicts AQI → logs data
├── test_model.py                               # Test one sample manually
├── requirements.txt                            # Python dependencies
└── README.md  

---

🚀 Features

✔️ End-to-End ML Pipeline

• Preprocessing
• Normalization
• Label encoding
• RandomForest classification
• Saved model + scaler for real-time inference

✔️ Real-Time Sensor Integration

ESP32 streams sensor values in CSV format:

CO,NO2,SO2,Ozone,Temp,Hum

Python listener:

• Normalizes input
• Predicts AQI category
• Prints live warnings
• Saves readings to my_live_dataset.csv

✔️ Build Your Own Dataset

Every real sensor reading is appended to:

data/live/my_live_dataset.csv

Use this for retraining later.

✔️ Modular, Clean Code

Arduino code, ML code, preprocessing, and live prediction are separated.

---

🧠 How the System Works

1️⃣ Preprocessing

Run:

python preprocess.py

This:

• Loads final_dataset.csv
• Cleans & normalizes data
• Categorizes AQI
• Saves processed dataset + scaler

---

2️⃣ Model Training

Run:

python ml/train_model.py

This:

• Loads processed dataset
• Trains RandomForest
• Saves model + label encoder

---

3️⃣ ESP32 Sensor Reading

Upload the code in:

arduino/esp32_mq135_dht/esp32_mq135_dht.ino

Reads:

• MQ135 analog gas values
• DHT11 humidity & temperature Then sends CSV lines through USB serial.

---

4️⃣ Real-Time Prediction

Run:

python sensor_predict.py

This script:

• Opens ESP32 serial port

• Reads CSV sensor data

• Normalizes using scaler.pkl

• Predicts using aqi_model.pkl

• Saves readings + predictions to:

  • data/live/my_live_dataset.csv
  • data/processed/live_readings.csv

• Prints warnings when category is dangerous

Example output:

2025-02-03 RAW=3.432,1.981,1.123,4.550,28.2,54.1 => AQI=Moderate

---

🛠️ Hardware Setup

🔌 Sensors

• ESP32 DevKit V1
• MQ135 (Air Quality Sensor)
• DHT11 (Temperature & Humidity)

🪛 Wiring

MQ135 → ESP32

MQ135 Pin	ESP32 Pin
AO	GPIO34
VCC	5V
GND	GND

DHT11 → ESP32

DHT Pin	ESP32 Pin
VCC	3.3V
DATA	GPIO4
GND	GND

---

🧪 Testing Without Hardware

Run:

python test_model.py

Outputs something like:

Predicted AQI Category: Unhealthy

---

🔧 Requirements

pandas
scikit-learn
joblib
pyserial

Install:

pip install -r requirements.txt

---

🎯 Future Improvements

• Better MQ135 calibration
• ThingSpeak/MQTT integration
• Mobile app dashboard
• Live charts

Vijeta Patel GitHub: https://github.com/Vijeta-Patel