Fronius MPP Hunting

mradtke
mradtke Solar Expert Posts: 53 ✭✭✭✭
Hello,

I have 2 strings of 7 180 watt panels in a somewhat south southwest facing array and a matching array facing south southeast. These 4 strings feed a Fronius IG 5100 inverter. There are no data logging options on the inverter. Under some conditions, I see the power reported on the Fronius display oscillating about 400 watts with about a 42 second period. At other times, the oscillation is not nearly as evident, but it always seems to be present.

I also noticed that the display updates about every 2 seconds. However, after 21 updates, it does a double update. Is it a coincidence that the display period is the same as the power fluctuation period?

Both the installer and Fronius aren't showing much interest in my problem. Does anyone have an idea of what's going on?

Thanks,
Mike -- Phoenix, AZ
«134

Comments

  • Solar Guppy
    Solar Guppy Solar Expert Posts: 1,989 ✭✭✭
    Re: Fronius MPP Hunting

    Well, a couple of thoughts,

    First, 7 panels in series for a sting inverter in AZ seem to small, I checked out the data sheet

    http://www.fronius-usa.com/worldwide/usa.solarelectronics/downloads/specifications_ig_4000_5000_4500lv_012008.pdf

    So you should verify 7 panels works in the heat. if your panels are 24V panels ( typical vmp of 34V ) it might work, if they are 18V panels ( typical for 180 watts ), the vmp point would be below the minimum needed for the inverter which is 150V

    Also, the Fronius inverters I tested had one very noticeable quirk, they change on demand the transformer windings "on-demand" depending on the input voltage. On the inverters I tested, its was @ 260V input that the switch would occur. If the Mppt is near this I could see some issues

    I'd run your panel configurations thru the Xantrex string calculator ( to get the array operational voltage ) and compare to what the IG5100 is spec'ed for. http://www.xantrex.com/support/gtsizing/index.asp?lang=eng
  • mradtke
    mradtke Solar Expert Posts: 53 ✭✭✭✭
    Re: Fronius MPP Hunting

    Hello Solar Guppy,

    Thanks for your comments.

    I have attached a text file which is really a dot.csv that you can load into about any spread sheet program. These data show the measurements that I took from the Fronius screen since I don't have the data logging option. From this you can see that the inverter is well above the 150 volt minimum during the oscillation.

    The inverter feeds one phase of a delta connected three phase service. This is a relatively rare type of service that looks standard in that there is the usual 120-0-120 feed, but then includes a "stringer" which measures 240 volts to either of the other lines. Of course, all three wires are 60 degrees out of phase. I mention this in case there may be some special consideration for this circuit.

    You'll notice from the data that the power pretty much follows a saw tooth with a period of 40+ seconds. A couple more things to note:

    1) The display updates about every 2 seconds. Every 21 updates it does a quick double update. That would be about 42 seconds, or the same rate as the power oscillation. Coincidence?

    2) It doesn't always oscillate. I thought that it was power or heat related, but for the last two days there have been no oscillations. Then today, it was already oscillating from the first time I observed it at about 08:00 and 1000 watts.

    3) Today it oscillated most all of today. At 13:50 is was bouncing between 2100 and 2700 watts. However, when I checked at 14:40, the power was constant at 2700 watts. Some passing clouds may have quenched the oscillation, but the more interesting thing is that at 14:40 the inverter was reporting the same value as the peak power an hour earlier. It hasn't started oscillating again, but there are some clouds now.

    Fronius is talking about doing some measurements, but it is over 10 days since I notified them of the problem.

    Thanks for your interest,
    Mike
  • Solar Guppy
    Solar Guppy Solar Expert Posts: 1,989 ✭✭✭
    Re: Fronius MPP Hunting

    unfortunately you will have to get much more accurate data before a manufacture would get some interest, or at least that has been my personal experience ... if your a home power subscriber ( Homepower issue 91 )

    http://www.homepower.com/article/?file=HP91_pg46_Cutler

    I tested the IG3000/4000 units in 2004 and found some quirks with them, which Fronius engineers admitted to, but hunting mppt was not one of the observations I measured

    I looked at the file, The voltage current relationship seem valid but without external data loggers one can never be sure what the unit is reporting is accurate.

    I'd continue pressing Fronius and see if the could hook-up and external logger OR swap out the unit, intermittent electronics can for sure make the Mppt track poorly.

    For sure, its not optimum performance and this is a great example to explain why I strongly recommend the Xantrex GT series inverters. Built in RS-232 communications and either GT-View of SG-View on the PC gives 1 second resolution for full performance and free data logging.

    If you can't get Fronius to own up to the issue, possibly you can get the install to install another manufactures unit
  • mradtke
    mradtke Solar Expert Posts: 53 ✭✭✭✭
    Re: Fronius MPP Hunting

    Hello Solar Guppy,

    Thanks for your additional comments and the link.

    I will continue pressing the installer and Fronius to get a resolution.

    The Fronius was chosen for the low voltage performance as you noted in your last reply. The physical constraints of the collectors make 28 the ideal number. According to the installer the choice was between a pair of SMAs or the Fronius.

    Today the system came up without oscillation. So, I looked if I could start the oscillation. This is what I came up with:

    Start the oscillation: Break the AC side and let the inverter cool for 15 minutes.
    Restart the inverter by re-establishing the AC connection.

    Stop the oscillation: Momentarily break the DC side.

    By the way, the average power when the inverter is not oscillating is several hundred watts higher than the peak power during the oscillations.

    I think that Fronius will take measurements tomorrow, but I think that this will only confirm what I already have reported. I'll follow up this post from time to time with progress reports. However, I am still open for advice.

    Thanks,
    Mike
  • autoxsteve
    autoxsteve Solar Expert Posts: 114 ✭✭✭✭
    Re: Fronius MPP Hunting

    I've got a very similar system to yours.

    I've noticed some 'dancing' with my inverter and called my installed (REC Solar) about it. I also spoke to an engineer from Fronius and the #s I had reported (I want to say around 200 watts or ~10%) at that time (back in March) were deemed 'normal'.

    I've had my system operating for a little over 6 months (3/17 'go live') and have produced almost 4.9 MWh to date.

    My thought on your sit is that the inverter is seeking the optimal point and it will dance a little. Have you had some high cloud / or overcast days to observe the behavior?

    How many kWh are you getting each day with your system?

    I'm getting about 23 kWh per day presently (over about 11.75 hours of operation).
  • boB
    boB Solar Expert Posts: 1,030 ✭✭✭✭
    Re: Fronius MPP Hunting

    It looked to me as though there wasn't really a trend up or down in his average power level (from his data). Just the sweeping ?

    But maybe I missed something.

    boB
  • mradtke
    mradtke Solar Expert Posts: 53 ✭✭✭✭
    Re: Fronius MPP Hunting

    Bob and Steve,

    Thanks for your replies.

    I don't understand your comment about a trend. My position was that since there was a peak power value and a lower power value in the 42 second oscillation, the MPP must be at least as high as the peak. I have learned to quench the oscillation by interrupting the DC (this doesn't always work the first time), I see that when the system is not oscillating, the inverter settles in at a stable MPP somewhat above the peak reported during the oscillation.

    Discounting any cloudy days, the power has been about 21 kWh / day for the last 2 weeks and the system has been oscillating. Today was the first day that I stopped the oscillation early on and monitored the situation to be sure that it didn't recur. The system produced 24 kWh today. That would be empirical proof that the oscillation is reducing the energy production.

    At the worst that I've observed, the difference between the maximum and minimum values that the system is oscillating between is over 1000 watts. That's not dancing. When the oscillation is quenched, the variation is a few 10s of watts. That's dancing.

    I have been exchanging email with Dan Fortson of Fronius USA. He told me that no one else has reported this problem. Is he the person that you contacted?

    Thanks,
    Mike
  • mradtke
    mradtke Solar Expert Posts: 53 ✭✭✭✭
    Re: Fronius MPP Hunting

    Hello,

    The installer came to my home and observed the hunting that I had described. Fronius backed up the visit by phone.

    One thing new that we learned is that the current, when measured with an external meter, appears to go to zero once per 42 second cycle. Fronius suggested the transformer tap switching might be a issue, but I was able to confirm that there was no audible click of the switch heard during the oscillation. Fronius has now suggested that we install a data logger on the system to gather more information ... something that I requested 2 weeks ago. It should be here in a couple of days.

    If you have a Fronius, I suggest that you go and look if you have this problem as well.

    I'll post again when I know more.

    Thanks for listening,
    Mike
  • niel
    niel Solar Expert Posts: 10,300 ✭✭✭✭
    Re: Fronius MPP Hunting

    good to know and keep us posted as to any developements.
  • autoxsteve
    autoxsteve Solar Expert Posts: 114 ✭✭✭✭
    Re: Fronius MPP Hunting

    I'll look for your update. You appear to have a different problem than I.
    The jumping around that I am experiencing is only a few watts and I don't see any jitter or changes in current like you describe (I watched my meter and it's smooth).
  • mradtke
    mradtke Solar Expert Posts: 53 ✭✭✭✭
    Re: Fronius MPP Hunting

    Hello,

    I received and installed the data logger today. The system began oscillating on startup. However, there was some cloudiness so I think that the data will be meaningless for today.

    The intent is to record at least one day of oscillations and then send the device back to Fronius for analysis.

    Thanks for listening,
    Mike
  • mradtke
    mradtke Solar Expert Posts: 53 ✭✭✭✭
    Re: Fronius MPP Hunting

    Hello,

    I am data logging today, but we have high cloudiness. The inverter is doing its 42 second oscillation though. The forecast looks like the next week will be cloudy.

    This gives me time to think. Since I have lost confidence in Fronius, how can I be sure the repaired or replaced unit is working properly? Is there some acceptance test that I can use to verify that it is operating at maximum power? What do you folks do?

    Thanks,
    Mike
  • boB
    boB Solar Expert Posts: 1,030 ✭✭✭✭
    Re: Fronius MPP Hunting

    I wonder what that Fronius would do if you were to power its PV input off of a suitable but steady power supply ?

    boB
  • autoxsteve
    autoxsteve Solar Expert Posts: 114 ✭✭✭✭
    Re: Fronius MPP Hunting

    If you wish, I can video record my fronius inverter output and meter output and email the files to you... let me know (pt or email)....
  • mradtke
    mradtke Solar Expert Posts: 53 ✭✭✭✭
    Re: Fronius MPP Hunting

    Hello,

    Thanks again for your comments.

    The data logging is over. I had 2 completely cloudless days and the inverted hunted the entire day both days. So, I sent the logging equipment back.

    Steve, I think that I know what the unit looks like when it is working properly, but thanks for your video offer. I used to be able to jolt the system into that state, Then, its output went up and it only hunted around a few 10s of watts. The hunting was random as well, not with the 42 second cycle.

    boB, I've thought about running the inverter from a different power source as well. However, the trigger may be due to some characteristic of my array, so it was probably a good call to do the data logging on site. I don't have a lot of confidence that they logged the right things though since the Fronius engineer mentioned a 5 minute logging rate. That may not do a lot of good with a 42 second oscillation in output. But, I'm giving them the benefit of the doubt.

    I am still concerned that even if Fronius makes a repair or replacement, how will I know that the system is operating at the maximum power point? I used to be primed to trust them that they got it right, but now that I see that they got it wrong, how can I verify that they get it right next time? What do you folks do the verify your systems are running at the maximum power possible?

    Thanks,
    Mike
  • Solar Guppy
    Solar Guppy Solar Expert Posts: 1,989 ✭✭✭
    Re: Fronius MPP Hunting

    If they don't come up with a resolution, get the installer to swap the Fronius with a Xantrex GT ... they are proven to work and Outsell the Fronuis 10 to 1 in the US last I knew the numbers
  • mradtke
    mradtke Solar Expert Posts: 53 ✭✭✭✭
    Re: Fronius MPP Hunting

    Hi Solar Guppy,

    Thanks for your comments, but the Fronius was chosen because of the panel configuration. I think that the Xantrex units don't match the panel configuration as well.

    No one has commented on a method that might be used to verify that their inverter is operating at the maximum power point. Do you folks just trust the inverter manufacturer? That's what I did until I saw the oscillation. Now I see things differently.

    I have devised a test that is pretty straight forward and should work with any system. The recipe is only about 4 or 5 steps long, but the theory is pretty complicated. I wouldn't ask anyone to trust the method until some experts have sorted through the theory with me. My point is that while the test is simple, the post to describe it would be pretty long.

    What is the list protocol on sticking to the title of the chain? Should I start a new chain with the description of my test method and theory behind it?

    By the way, I haven't heard anything more from Fronuis yet.

    Thanks,
    Mke
  • BB.
    BB. Super Moderators, Administrators Posts: 33,431 admin
    Re: Fronius MPP Hunting

    Since this is "your thread"--it would be fine to include your discussion here. Keeps everything together in one place--easier to find / follow for others later on.

    If the detailed post is "long"--you might want to attach it as a file attachment.

    -Bill
    Near San Francisco California: 3.5kWatt Grid Tied Solar power system+small backup genset
  • boB
    boB Solar Expert Posts: 1,030 ✭✭✭✭
    Re: Fronius MPP Hunting

    One thing you can do as kind of a sanity check is to measure the open circuit voltage of the array. If there is NO shading of the array, the MPP voltage should be roughly 75% of that voltage. 75 to 80 % of Voc in full sun and no shading on the array.

    boB
  • mradtke
    mradtke Solar Expert Posts: 53 ✭✭✭✭
    Re: Fronius MPP Hunting

    B.B: Thanks for the advice.

    boB: Good plan .. thanks.

    OK, so here is my proposed method of determining whether an inverter is operating at the system's maximum power point. First the recipe:

    1) On a sunny day record DC voltage and current measurements. Do this for at least long enough for the inverter to hunt around a few times for its MPP and for a few tens of measurements. There is no need to keep the measurements in pairs, but do have the same number of voltage and current measurements.

    2) you may wish to use a spread sheet for the rest of this, but it's easy to do manually as well. Make columns of the measurements and sort the voltage measurements in ascending order and the current measurements in descending order.

    3) Calculate a third column as the product of the adjacent voltage and current measurements. This is the power represented by each measurement pair.

    4) Ignore the current column, and plot each power and voltage pair as a scatter diagram with power on the vertical axis.

    5) Examine the results. Some possibilities:

    a) Dots form some sort of a peak. The highest value is the MPP and the inverter is walking back and forth across it as expected. There may even be a heavy concentration of dots in the middle if the inverter is spending most of its time at the MPP.

    b) Dots form some curve without a peak. The inverter is never finding the best spot. If the line looks very horizontal, perhaps the data isn't accurate enough, or the scale needs to be expanded to identify the peak.

    I measured my malfunctioning Fronius for about 5 minutes and got 48 voltage and 48 current measurements. My plot was a diagonal line with no peak. The heaviest concentration of points was about 200 watts below the highest end of the line.

    The theory starts with some assumptions:

    a) The solar panels have consistent light during the test.

    b) In the small range that the inverter hunts around for the MPP during the measurement period, the solar panels may be modeled to be a fixed voltage source in series with a fixed resistance.

    c) The current and voltage measurements do not have to be especially accurate, but they must be repeatable. Therefore, the inverter's own display may be used to take the measurements.

    Simultaneous current and voltage measurements would be perfect. However, those values are probably not possible to get. The measurements may be sorted, combined, and treated as simultaneous because of assumption #b. In other words, the highest voltage must be paired with the lowest current and so on.

    The calculated power plotted against voltage should represent a sample of the inverter's attempts to maximize power by trying different voltages. Because of assumption #b, this may be considered equivalent to the inverter trying different currents. The plot could just as well be current versus power as well. The important thing is whether the plot shows a peak in power and whether the inverter spends most of its time at the peak.

    So, what do you think?

    Thanks,
    Mike
  • Solar Guppy
    Solar Guppy Solar Expert Posts: 1,989 ✭✭✭
    Re: Fronius MPP Hunting

    You don't have the equipment to make measurements in the resolution necessary to have meaningful data for pin-pointing the vmp. You need a quality power analyzer that does simultaneous measurements and handles power factor as well ( both on the solar and AC sides )

    Typical large inverters probably hunt in the 10-20 watt range

    Instead, you should focus the effort in make Fronius get more involved

    First, insist they change out the inverter as its hard to believe it needs a 200 watt perturb window to find the mppt value and stay on it, you just likely have a bum inverter. There are many ways internally a unit can have intermittent electronics that would cause the ( current sense transformer, faulty op-amps, for example )

    Unless you want to spend a couple of grand renting the equipment needed ( Yokogawa WT2030 and a curve tracer ) its not a worth while for your time, spend that getting Fronius to resolve your issue

    200 watt mppt perturb is not normal and not what I saw when I tested a IG2500/IG3000 in 2004.

    Issues I don't like about this manufactures:

    -> limited voltage range

    -> active switching of the transformer primary for even the limited voltage range, very odd design

    -> efficiency is different for each of the 3 ranges

    -> Reported power is greater than 5% of actual power produced

    -> Small heat sink that requires constant fan cooling ( about 1/20th the size of the Xantrex GT heat sink )

    But in the month of testing I did, I never saw any large perturb issues like your seeing.

    Again, get Fronius to swap out the unit as the next step
  • mradtke
    mradtke Solar Expert Posts: 53 ✭✭✭✭
    Re: Fronius MPP Hunting

    Solar Guppy,

    Thanks for your detailed response.

    Since there is no way to get simultaneous measurements there is always the dependency of getting both the current and voltage measurements faster than the inverter can change them. That's why I cooked up my scheme which did not depend on simultaneity. The Fronius internal metering resolution is 1 volt and 0.1 amps which should be just good enough to verify a successful hunt over a few tens of watts. I think that if my proposed method shows a peak, it will be proof of finding the maximum power point. If it doesn't show a peak, not much can be concluded. On the other hand, I respect your experience in these matters and my hopes aren't very high for my proposed measurement. Which leaves us back at the point of trusting the inverter manufacturer unless some pretty sophisticated equipment is used.

    You mentioned power factor and thus implied inverter efficiency verification as well. That is a whole new can of worms, isn't it?

    So, do you ever see a partial inverter failure such as I have, or has your experience been with total failures? If you are seeing partial failures, how do you find them? I guess that I am back to some sort of system verification test.

    Fronius has been dragging there feet on this. I am keeping the pressure up, but I am also trying to be cooperative so that they might learn something from this failure. Last week their excuse was the Solar show in San Diego. This week they say the their "brass" is in town and taking up their time.

    Thanks,
    Mike
  • Solar Guppy
    Solar Guppy Solar Expert Posts: 1,989 ✭✭✭
    Re: Fronius MPP Hunting

    I'm surprised Fronius is not being responsive ...

    When I did my work and had questions, I literally got responses same day. I would think the same tactics that one would use for any company when not getting the support needed would apply, like getting the manager or directors email address and moving it up the chain.

    You should be focusing on this, getting the proper person at Fronius to swap out your unit. unfortunately, my contacts from 2004 have moved onto other company's so I don't have emails to offer ... I'd call the Michigan office and not quit until I got a hold of the top person there and had there commitment to resolve the issue

    I know where your at ... read my HP 91 article and my journey in inverter performance. Keep in mind, you need Fronius to step up here and better data collection is not a job if the customer.

    You might want to bring this thread to there attention ... I suspect that would help in getting the support you need.
  • crewzer
    crewzer Registered Users, Solar Expert Posts: 1,832 ✭✭✭✭
    Re: Fronius MPP Hunting

    Mike,

    Perhaps I missed this info... Which 180 W modules (manufacturer and model) are you using?

    Regards,
    Jim / crewzer
  • mradtke
    mradtke Solar Expert Posts: 53 ✭✭✭✭
    Re: Fronius MPP Hunting

    Jim: 28 Suntech 180 watt panels in 4 strings.

    Solar Guppy: Thanks for your comments. I am dealing with Michigan.

    Fronius got back to me about the data that I captured for them with their data logger. According to them, the collected data described a perfectly functioning inverter. I reminded them that I was concerned about that since before they sent me the logger, they told me that it logged at a 5 minute rate. My oscillation has a 42 second period.

    Yesterday I did another measurement on my system with Fronius on the phone. We started out by my reading off current and voltage measurements to them for a while. They got lots of good data points on the hunting. Then they gave me the crack to enter a service mode on the inverter and had me change the inverter from MPP to fixed voltage mode. We chose a voltage inside the range that it had been hunting and the inverter ran fine this way and did not hunt. We then tried a voltage just below the lowest voltage seen during the hunt, but the inverter refused to accept this voltage setting. The error implied that the setting was too far below the current open circuit voltage value. I'm not sure that much was learned from this experiment, but Fronius has decided to send out a new inverter.

    I still would like to know just what the failure is and I am a little concerned that a replacement inverter may fail in the same way. Some thoughts:

    1) At first, constant voltage mode seemed like the ideal way to verify that MPP is being found. However, the fact that the fixed voltage could not be set low enough to overshoot or even reach the MPP makes it useless for this purpose. Or, is this a hint of what's wrong with this inverter?

    2) Doing these measurements yet again caused me to notice that the lowest indicated voltage was always 232 volts, and just before the current dropped to zero. So, yesterday afternoon, when the power output was much lower than it had been when I made all of my other measurements, I watched the voltage again. Just like before, as the system oscillated. The voltage slowly dropped to 232 volts, then the current dropped to zero, the voltage shot up to the open circuit value, and the cycle repeated. Is the 232 volts another hint about the cause of the failure?

    I'll post again when the new inverter gets installed. You know that I'll be running measurements again.

    Mike
  • BB.
    BB. Super Moderators, Administrators Posts: 33,431 admin
    Re: Fronius MPP Hunting

    What is the minimum Solar Panel input voltage rating of the inverter?

    -Bill
    Near San Francisco California: 3.5kWatt Grid Tied Solar power system+small backup genset
  • mradtke
    mradtke Solar Expert Posts: 53 ✭✭✭✭
    Re: Fronius MPP Hunting

    Bill,

    Thanks for your question.

    From the manual: MPP-voltage range: 150-400 V.

    This inverter was chosen specifically for its low operating voltage.

    Mike
  • BB.
    BB. Super Moderators, Administrators Posts: 33,431 admin
    Re: Fronius MPP Hunting

    Well--there went my guess...

    -Bill
    Near San Francisco California: 3.5kWatt Grid Tied Solar power system+small backup genset
  • Solar Guppy
    Solar Guppy Solar Expert Posts: 1,989 ✭✭✭
    Re: Fronius MPP Hunting

    I'm pretty sure 232VDC is very close to the transformer primary switch voltage trigger for their units .. I seem to recall it was a 230 - 260V window, its probably a bit different for the 5kw unit but close.

    Fronius has a very odd way to extend the voltage range, they have the isolation/ step up transformer ( between the DC-DC and DC-AC converters ) with three taps. The unit via relays under software control will switch the tap to keep the secondary voltages down, a really big issue with 2 stage units.

    They are the only manufacture to do this. The unit in theory has enough hysteresis to not oscillate but in your case I have a hunch why your seeing this and other sites don't

    If the new unit performs the same, my educated guess is the root cause is you vmp is to close to the software trigger switch combined with you have split facing arrays, the vmp between the SE and SW arrays is probably make a very flat and even possibly double hump vmp point .. so the unit keeps swinging between the voltages and keeps triggering the hysteresis algorithm which requires the unit momentarily stop generating power

    You can prove this by disconnecting one of the arrays ( just run on the SE or SW arrays ) and see if the hunting stops. If this is the case one fix would to be rewire the arrays so half come from each SE/SW ( 7 from SE string then would goto 7 of the SW panels ).

    Its unfortunate that you probably have found a condition there units are not able to work well with .. I'll keep my fingers crossed that I'm wrong

    By the way, the Xantrex GT5.0 works fine with split arrays as does the Sunnyboys ... and if you gave up one panel you could have had three strings of 9 panels for either a GT5.0 or a SB5000U both are much better units.

    Again, if the new Fronius does the same thing, get the installer to give you either a GT5.0 or SB5000U, its not YOUR fault the inveter doesn't function correctly. Oh and if you haven't done so, get Michigan to read this post and send of to Austria ... I bet once the real engineers read this you'll be having visitors with lots of test equipment

    Oh one last thing, since you can put the unit into fixed vmp mode, it would be simple to start at the top of the range and keep setting lower voltages, I'm sure you can find the vmp this way and if the unit refuses to run a a commanded voltage, its may be below the transformer switching point or there 2 stage design can't operate below the vmp point, common for two stage units .. try 5 volt steps and then smaller are you hone in on the vmp .. should take more than a minute to do
  • BB.
    BB. Super Moderators, Administrators Posts: 33,431 admin
    Re: Fronius MPP Hunting

    Another possible experiment/fix to try (for your installer, if you have room)--try adding a panel or two to each string and raise the working voltage (even if temporally) out of the "instability zone".

    -Bill
    Near San Francisco California: 3.5kWatt Grid Tied Solar power system+small backup genset