Laboratory test of Deye SUN-3K-SG04LP1 hybrid inverter (+ YouTube video)

[Jiri Bekr]

Hi,
I have done a comprehensive test unter controlled laboratory conditions of this Deye ongrid hybrid inverter - SUN-3K-SG04LP1-24-EU
It is a 3kW 24V / 230V@50Hz hybrid inverter. The data like efficiency, rules for efficiency gain, power quality, etc. should be applicable to all hybrids in the SG04LP1 - 3kW to 6,2kW. It may as well be appliacable to SolArk inverters (made as OEM/ODM by Deye for North America)

Test is published in YouTube video:
https://youtu.be/KdgEwOCb_wg
I am also gonna publish the test in text form here in this post, incl. all data, tables and graphs.
Please apologize for the tables and some graph labels being in Czech. Most of the stuff is self-explanating anyways, so I don't think there's a need for translated spreadsheet.


Test setup:

BATTERY PACK
8S 1P DeJin 25,6V 314Ah. BMS: JKBMS B2A8S20P-CDHR. COMM: Deye CAN protocol.
Battery - Deye path: cables 2x 25mm², length cca 1,2m, 2 passes through ferrite core. Current measured by Amprobe AM-220-EUR in inverter tests, or by Hioki CM4375-50 in other tests. Current ripple measured by Hantek 650A hall sensor current probe + Owon HDS25S oscilloscope. Voltage measured directly at Deye's input terminal bolts by Hioki DT4282.

INVERTER TEST
Output V, I, P measured by Atorch socket tester. Output THD-U measured by Hioki CM4375-50 up to 30th harmonic. DC offset measured by UNI-T UT61E+ (beware, this data is unreliable, I have to have a look at it). Load was done by perfectly resistive stuff - classic wolfram/tungsten bulbs. I have an army of 200W, few 300W, few 500W, and some smaller ones. Needless to say that being in the same room with 3,6kW in incandescent bulbs is quite an experience...

LONG TERM LOAD
Deye was loaded by 2kW space heater and 1kW of bulbs for 2 hours. ThermalMaster Thor 002 thermal camera was then used to evaluate heating up.

MPPT
Input I measaured directly by UNI-T UT61E+ (plus bypass switch to prevent shunt overheat). Input V measured cca 15cm before inpuit MC4 terminals to the Deye by Zoyi ZT-703S. I used 6S and mostly 8S panels: 370Wp 34,4Vmpp 10,76Impp.

GRID CHARGING
Vevor 24V 3,5kW hybrid inverter served as a source for so-called GRID (as I cannot legally connect ongrid hybrid inverter into my utility in my country without load of paperwork). Input U, I, P, S, cosPHI was measured by Atorch socket tester. GRID charger in the Deye was being set between 5A to 140A.

UPS TEST
Vevor 24V 3,5kW inverter served as the GRID source. Then there was my BlackoutBox, and then the Deye. BlackoutBox is really just a box with 2-pole switch, one pole switches the line voltage (from plug to socket), and second pole switches 6 volts from integrated batteries to output banana plugs, that serve as trigger signal. Oscilloscope was Fnirsi 1013D touchscreen tabletoid "scope". CH1 is voltage from LOAD output of Deye attenuated by 100:1 probe. CH2 had the trigger 6V signal from BlackoutBox.
 

Test results:

INVERTER TEST
Efficiency is nice, around 93% between 900W-2,9kW. Efficiency doesn't fade with high power (components are being driven within their optimal parameters), power quality is perfect at 0-120% load, and the inverter handles 20% overload long-term without complaining or bad smells. See the attachment.

LONG TERM LOAD
After 2hrs on 100% load, the chassis temp and heatsink temp were nicely low, which tells about the efficiency and cooling capacity. Only the inductors were hotter, which is normal. Note the electrical tape on the heatsink for emissivity. See the attachment.

MPPT
The MPPT reached top 94,1% overall efficiency, or 95,85% of corrected efficiency (correction for inverter zero-load idle consumption). Extremely beautiful numbers. Efficiency rises with string voltage, and falls with PV current. Unfortunately, the low voltage PV limit is too high, hybrid inverters usually begin at 40-60V, Deye begins at 150V.

GRID CHARGING
The Deye inverter surely has a powerful grid charger, 2x more powerful compared to cheaper alternatives (Deye 140A, others usually 80A). The Deye surely has active PFC, maybe through an intelligently controlled synchronous rectification? It reaches a perfect power factor at 30-40A already, and then keeps PF 0,99-1,00 to the end (140A). The charger efficiency is slightly on the worse side, generally around 90%, but grid charging is rather a complementary function, and not a key function, so that's normal.

UPS TEST
Average transfer time was 8,1ms, fastest 6,6ms, slowest 11ms. Slowest reactions were in places when the voltage was approaching zero-cross situation (near 0°and 180°), as in those areas, blackout can be misidentified at the beginning as a transient or bad power quality. Fastest reactions were at higher absolute voltages (near tops and bottoms of sine wave), where it is apparent that it's a blackout. The inverter has a phase locked loop, so it always continued in phase with the former (now lost) GRID sine wave. This means it's safe for compressors and motors.


I can only attach 10 files at 8MB total (understandable, Dave prolly doesn't have petabytes of harddisks). So I am as well attaching a link to all data, UPS scope screenshots (incl. zoomed-in with cursors), and thermal images.

All data:
https://bit.ly/GDriveLiFePO4Data
Please note, I use this shared folder for all LiFePO4 batteries, hybrid inverters, car inverters, and other PV stuff that I test. You need to enter the Deye folder.

Hope this test helps as many people as possible.