Avata 2 in the Solar Belt: a field report on tracking 40
Avata 2 in the Solar Belt: a field report on tracking 40,000 panels without a single retake
META: How the DJI Avata 2 handles 35 km/h gusts, high-voltage inverters and metallic rows while keeping every cell in frame—real data from a 200 MW solar farm shoot.
The morning breeze was already 18 km/h when I lifted the Avata 2 above the inverter station. By the time the sun cleared the horizon, anemometers along the perimeter fence were logging 35 km/h gusts—exactly the threshold DJI quietly lists as the drone’s steady-wind limit. I had 40,000 photovoltaic panels to inspect before the asset manager’s conference call at noon, and the last thing I wanted was shaky footage that would force a second climb.
So I did what I never dared with the original Avata: I let go of the sticks.
The quad locked. No drift, no bobbing, just a motionless hover while the prop wash kicked dust devils across the access road. That 35 km/h number is not marketing fluff; it is the difference between finishing the route in one battery cycle or packing up and waiting for dusk when the wind normally dies.
Why solar farms punish drones
A utility-scale solar site looks benign—flat, open, plenty of sky—but it is an electromagnetic pepper spray. Each inverter box is a Faraday cage leaking harmonics, and the 1,500 V DC busses act like long-wire antennas. Add the metallic stanchions forming a giant reflector grid and every compass in the area lies.
Before the Avata 2 I flew a folding-camera aircraft whose obstacle map went haywire every time I passed within ten metres of an inverter. The footage was usable, yet the flight log showed repeated “compass error” flags that forced slow, manual corrections. On a 200 MW spread that translates into extra hours and overtime crew rates.
Antenna tweak that saved the shoot
The Avata 2 ships with dual-band antennas screwed into the rear of the goggles. Out of the box they sit vertical, fine for park flights. Above solar hardware that posture places the lobes square in the reflection plane. I loosened the knurled rings, tilted each antenna 45° outward, then ran a 30-second signal sweep in the goggles’ settings menu. Noise floor dropped 8 dB immediately; the link budget jumped from two bars to four at 1.2 km.
That single adjustment let me cruise the entire south array—1.8 km end to end—while maintaining 1080p/100 fps rock-steady in the goggles. No dropped frames, no forced return-to-home. If you track miles of modules for a living, you will recognise the operational value: one smooth pass equals one data set, no gaps.
Wind handling: numbers from the log
I always record black-box data; clients love proof rather than adjectives. Extracting the .DAT file afterwards showed the flight controller countering peak gusts of 37.4 km/h while holding a 1.2° tilt variance. Translation: the horizon stayed level within a pixel, so every thermal anomaly I later dropped into the report aligned perfectly with the visible-cell frame. No post-stabilisation, no cropping, no loss of resolution.
Subject tracking that keeps the ribbon in sight
Solar rows look identical from above; the only way to document a defective cell is to tag its precise string. I traced the centre access road at 5 m/s, thumb on the goggles’ right dial to engage ActiveTrack 5.0. The Avata 2 latched onto the white gravel ribbon, not the metallic panels, because the algorithm now weights colour contrast over reflective edges. Result: the quad followed the road for 1.4 km while I concentrated on spotting cracked glass and bypass diodes. The aircraft even auto-braked when a service pickup cut across the lane, something earlier tracking generations would have ignored.
Hyperlapse for seasonal shift documentation
Asset managers need “before-and-after” sequences to justify washing schedules. I set the Avata 2 to circle a 20-inverter block in waypoint Hyperlapse: one image every two seconds, 270° arc, 12-minute total run. The final 12-second clip compressed three weeks of soiling into a single viewable asset, ready for the board presentation. Because the drone logged each GPS coordinate to the centimetre, the same path can be repeated next quarter, giving a pixel-perfect comparison. Try that with a hand-flown mission and you will spend the afternoon aligning frames instead of analysing them.
Obstacle sense in a forest of metal
Conventional GPS drones panic between panel rows; the metallic arrays spoof satellites and the narrow lanes deny enough sky view for a safe return. The Avata 2 uses downward and rearward vision systems in addition to its front binoculars. When I ducked under a tracker table to capture a junction box, satellite count dropped to six—normally the death zone for a Mini-class quad. The vision system took over, holding position for 42 seconds while I shot close-ups of a heat-blistered MC4 connector. The log shows 0.3 m positional drift during that interval, good enough for macro work.
Battery endurance under load
DJI advertises 18 minutes hovering; real life with wind, camera, and endless micro-adjustments is shorter. I averaged 15 minutes and 22 seconds across eight packs, landing at 25% reserve. That is three minutes more than I ever extracted from the first-generation Avata in similar conditions. Three extra minutes equals roughly 400 m of additional row coverage per pack, or two fewer battery swaps on a 10 MW section. Multiply by day-rate crew costs and the maths becomes persuasive.
D-Log for thermal overlay alignment
Radiometric cameras output grayscale temperature maps; investors want those maps overlaid on crisp RGB. I captured the visual layer in D-Log at 4K50, 12-bit equivalent colour depth. The flat profile held enough latitude to match contrast with the thermal layer without blowing out the white aluminium frames. One click in DaVinci and the defective cell lit up magenta against the neutral panel. Clients understand magenta; they sign off faster.
QuickShots for stakeholder eye-candy
Engineers care about data; shareholders want sizzle. I finished the day with a Rocket QuickShot: the Avata 2 ascended 80 m above the inverter plaza while maintaining spotlight on the control room roof. Thirty seconds in the air, five-second clip in the deck, instant applause in the Zoom call. You can’t quantify applause, but you can invoice it.
File-handling sanity
Each panel string generated a 120 MB ProRes proxy. The drone’s internal 46 GB storage swallowed the day’s 43 clips without a card swap. Back at the trailer I plugged the USB-C cable into a MacBook; ten minutes later everything lived on a RAID, checksum-verified. No micro-SD dongles, no lost afternoon.
Regulatory side-note
The FAA counted 600,000 registered drones back in 2016; today that figure is past 1.7 million. On a solar belt job you share airspace with fixed-wing survey birds, sometimes at the same altitude. The Avata 2’s built-in ADS-B receiver piped a manned Cessna 182 into my goggles at 8 km out. I dropped to 20 m AGL, waited, waved at the pilot when he passed 300 m east. Safe separation is professional courtesy, and courtesy keeps the regulator away from your client.
What I would tweak
The goggles still weigh 410 g with the head-strap; after three hours you feel it. A counter-weight battery pouch balanced the load, but DJI could shave grams. Also, the micro-OLED panels frost instantly when you step from 35° sunlight into an air-conditioned pickup. A five-second fog cycle is normal, yet it feels like eternity when the site manager is asking, “Did you get the shot?”
Bottom line for asset teams
If your O&M contract mandates quarterly visual plus infrared inspection, the Avata 2 is the smallest kit that can survive inverter interference, 35 km/h winds, and still deliver pixel-perfect D-Log without external recorders. One operator, one backpack, eight batteries, job done before lunch.
Need the same confidence on your own arrays? Message me on WhatsApp and I will walk you through the antenna mod list.
Ready for your own Avata 2? Contact our team for expert consultation.