Project: a smart plant watering system

What you will build and what you need

You will build a controller that waters one or two pots. It measures soil moisture every ten minutes, runs a pump for a few seconds if the soil is too dry, and publishes the moisture level and pump activity over MQTT. It keeps working even if Wi-Fi is down, because the decision happens on the device.

Design

Keep all electronics above and away from the water, and mount the board in a box with cable entry from below so drips cannot run into it. Use only low-voltage parts; never build this with a mains-powered pump unless you are qualified. The GPIO drives the MOSFET gate, the MOSFET switches the pump's negative lead, and the diode sits across the pump terminals.

Build and calibrate

1. Connect the moisture sensor to 3.3 V, ground and GPIO 34. Print raw readings in dry air and in a glass of water.
2. Push the probe into dry soil and note the value (DRY), then water thoroughly and note it again (WET). Capacitive probes read higher when drier.
3. Wire the pump through the MOSFET and test it for one second from the REPL.
4. Upload the code, choose a threshold suitable for your plant, and watch it for a few days before trusting it.

import time
from machine import ADC, Pin

soil = ADC(Pin(34)); soil.atten(ADC.ATTN_11DB)
pump = Pin(26, Pin.OUT, value=0)       # pump off at boot
DRY, WET = 2600, 1450                  # from your calibration
THRESHOLD = 35                         # percent
RUN_S = 5                              # seconds per watering
MAX_PER_DAY = 6                        # hard safety limit
MIN_GAP_S = 3600                       # wait for water to soak in

runs_today, last_run, day_start = 0, 0, time.time()

def moisture_pct():
    raw = sum(soil.read() for _ in range(16)) // 16
    pct = (DRY - raw) * 100 // (DRY - WET)
    return max(0, min(100, pct))

def water():
    pump.value(1)
    try:
        time.sleep(RUN_S)
    finally:
        pump.value(0)                  # always switch off, even on error

while True:
    now = time.time()
    if now - day_start > 86400:
        runs_today, day_start = 0, now
    m = moisture_pct()
    if m < THRESHOLD and runs_today < MAX_PER_DAY and now - last_run > MIN_GAP_S:
        water()
        runs_today, last_run = runs_today + 1, now
    # publish m and runs_today over MQTT here (see the temperature monitor project)
    time.sleep(600)

Safety and security checklist

• Fail safe: pump off at boot, off in a finally block, and a hard daily limit in firmware.
• Remote commands are requests: if you add an MQTT "water now" command, the device must still apply its own limits.
• Authenticate commands: only your controller account may publish to the command topic; sensors and dashboards are read-only.
• Detect faults: alert if moisture does not rise after watering (empty reservoir, blocked tube) or if readings jump to 0 or 100 per cent (sensor fault).
• Electrical safety: low voltage only, drip loops on cables and a fuse on the pump supply.
• Usual device hygiene: TLS, unique credentials, IoT VLAN, no debug services left on.

Tip: this project shows an important IoT principle in miniature. Any actuator that can cause physical harm must enforce its safety limits locally, so that a network attacker, a software bug or a stuck message cannot drive it beyond safe bounds.