Status: work in progress. Entry for the Nordic "Sustain the System" contest. Build planned Nov 2026 - Apr 2027.
The problem
In East Java, dragon-fruit farmers light every plant at night to make it flower out of season. A typical 25 x 100 m plot has 6-7 rows of 44 plants with a 9-15 W LED lamp each: about 300 lamps and 3-4 kW, on every night from February to September. These systems fail silently at night. When many farms switch on together the grid sags to 135-150 V. Cheap 220 V bulbs then dim, die or draw more current, which burns cables and magnetic contactors: on plots far from the transformer, farmers replace the contactor 2-3 times in 20 nights. 45 V DC systems are safer, but they carry 10+ A per row, so a loose joint heats up until it burns. A dark row means a part of the field that doesn't flower.
What BloomGuard does
A Bluetooth monitor in the plot's panel (nRF54LM20 DK) plus one far-end node per plot. It measures grid voltage, per-row current, power factor, leakage current, temperatures, humidity and light. It learns each row's healthy power-vs-voltage curve, so it can estimate how many lamps a row has lost even while the grid voltage swings. It also tracks joint resistance on DC rows and SSR/switch stress, and warns days ahead. Rows are switched with zero-cross SSRs instead of contactors: staggered on, dimmed during sags, tripped on dangerous leakage. Status is broadcast over Bluetooth LE (Long Range / Mesh) to the farmer's phone, with no internet needed.
Plan
Nov-Jan: bench rig with real AC bulbs, DC lamps, PSU, contactor and SSRs, with a variac sweeping 135-230 V. Induce faults and train the models.
Feb-Mar: install on real AC and DC plots in Banyuwangi during the flowering season.
Apr: full build guide, code, schematics and PCB files.
Full BOM (beyond the list above): zero-sequence CT for leakage current, ACS758 Hall sensors or INA226 shunts for the DC rows, SHT31 humidity sensor, rain sensor, BH1750 light sensors, MOSFET PWM dimming stage, MCBs per row group, IP65 enclosure.







Comments