ponyo19 / ForecastNet

ML-enhanced weather station network using ESP32, Raspberry Pi, MQTT, LAMP stack, and tinyML

ML-enhanced weather station network

This project is a IoT weather station network that displays past data and forecasts.

Overview

• A station in the network consists of an ESP32 and a BMP280 sensor.
• The Raspberry Pi works as an always-on server.
• MQTT is used for communication between the stations and the server.
• LAMP stack (Linux/ Apache/ MySQL/ PHP) and Python manages and stores data in a MySQL database and builds web interface for displaying data.
• Prediction is performed using tinyML.

Preparation

Parts required

• A Raspberry Pi board.
• An ESP32 and a BMP280 sensor for each station.

Required libraries

For ESP32:

• esp32 by Espressif Systems
• Adafruit Unified Sensor
• Adafruit BMP280
• NPTClient by Taranais
• PubSubClient by Nick O'Leary
• ESPAsyncWebServer
• AsyncTCP

For Raspberry Pi Raspbian OS:

• apache2
• php
• python3
• mariadb-server
• php-mysql
• phpmyadmin
• mosquitto
• mosquitto-clients

How to run this project

On ESP32

• Setup the board following this schema
• Install all required libraries
• Load the source code at ./esp32/esp32_sketch/esp32_sketch.ino to Arduino IDE
• Change WiFi and MQTT credentials
• Upload the modified code to ESP32

On Raspberry Pi

• Install all required libraries
• Setup mosquitto MQTT broker and MySQL database following this tutorial
• EXTRA: Check if the broker port 1883 on the server is open by running nmap -p 1883 <broker address> from a computer in the local network. Open the port if closed.
• (Recommended) Create and use a Python virtual environment
• Run pip install -r ./raspberrypi/requirements.txt to install necessary Python libraries
• Modify MQTT and MySQL credentials and run python ./raspberrypi/forecastnet_server.py to start the server for uploading sensor data.
• Data sent from ESP32 can now be browsed by accessing <your server address>/myphpadmin from a browser on any computer in the local network

... To be updated