I've been seeing a lot about this 100Ah / 1kW rule of thumb for GTI's having battery banks. I'd like to tell about the project I'm working on, and I would appreciate it if somebody would explain the necessity of using 600Ah of battery with the XW6048.

We are developing a 5kW fuel-cell power supply which needs to run continuously. Rather than load it with a $15k load bank and do nothing more than toss the energy away, we would like to tie the output into either our 120VAC or 240VAC line here in the building (don't worry, we plan to bring an electrician in).

Thus, the XW6048, which is the only GTI I can find which can work with the DC voltage range we need (we can't build a multi-100V system), and it also has a dynamically settable output current limit, which will be useful for controlling the inverter's behavior for testing purposes. I've checked with Outback, but they don't have such a feature, at least not one that can be set dynamically (that's what customer support said anyways).

So, my questions:
(1) If I can guarantee 5kW of power into the GTI at all times from our power supply, do I really need 500Ah or more of battery capacity?
(2) Does anyone know of an analysis from which this 100Ah / 1kW figure comes from, or is this purely empirical and best-practice based?

****************************

The reason I ask these questions is that I've been trying to do a simple (simplistic?) analysis myself. From what I understand, the current input waveform of your typical true sine wave inverter is a sine-squared waveform, and the amplitude of the waveform is basically reflected back from the load. I know this from talking with the folks over at Zahn Inc., who make a lot of different converters. The way it was explained to me is on the principle of power-in = power-out. You take your peak output current (for a 2.5kW / 120VAC load, this would be sqrt(2) * 120V / (120V^2 / 2.5kW) ~=30A) and multiply it by your peak output voltage (sqrt(2) * 120V) over your worst-case (minimum) input voltage (say, 27V). This would be a peak input current of 185A, not including inefficiencies (example taken from http://www.zahninc.com/documents/IM220c.pdf). Ergo, the input current waveform to something like the XW6048 would probably not be terribly different.

Now, if I understand correctly, the battery bank attached to the XW6048 acts as a charge storage device, not only for backup power, but it also takes the brunt of the AC currents being pulled by the GTI. Thus, if the GTI is pulling a sine-squared current waveform of large magnitude, and I subtract from the waveform the current being sourced by our 5kW power supply, the resulting waveform should be the battery current waveform.

If I assume a constant 5kW, 44V in (worst-case), what I find is that the total current waveform pulled in should be Iin(t) = 227A * sin^2(2*PI*F*t), where F = 60Hz. This is a sinusoidal waveform having a 120Hz frequency, which sounds similar to what SG is talking about here: http://www.wind-sun.com/ForumVB/showthread.php?t=11709. (Post #7)

Next, at 5kW, 44V in, the DC current into the GTI is 5kW / 44V ~= 114A.

Thus, the battery current waveform should be Iin(t) = 227A * sin^2(2*PI*F*t) - 114A. If you integrate under this waveform, you get 0A, which is what you would expect for an AC battery current. Anything != 0 would imply a constant charge or discharge of the battery with some DC current.

Whew. Anyways, if you take the RMS of this current waveform, you get about 57ARMS, and a peak battery current (charge OR discharge) of ~114A.

The rest of the discussion assumes that I've not misunderstood something so far about how the GTI functions.

Now, what is the important factor here? Is it the repetitive > 100A charge and discharge of the battery bank, or is it the RMS current, which from what I understand is what actually generates the heat? The reason I ask is that most of the figures I run across for battery impedances are in the mOhm range. Yet even if my battery pack had an impedance of 50-mOhms, the RMS current would cause ~160W of heating in the packs. 160W dissipated over four packs doesn't sound like a lot for their size, but maybe it is? Possibly more important would be the voltage swing of the bank: 50-mOhm * 57ARMS = 2.85VRMS, which may be too much of a swing for a typical GTI.

What I'd like to know is if my analysis relates in any way to the real world of battery bank sizing. Obviously, a higher capacity bank would imply a lower battery impedance (right?), which would mitigate heating and voltage ripple. I am thinking that the periodic high-current charge / discharge plays a role also, in particular because there are limits to charge rates which are healthy for the bank (C/8?). If I want to ripple-charge the bank with a 114A-peak current every half-cycle there might be some issues beyond heating.

I'm kinda new to this stuff, so please be patient.

Regards,
Tele

I'm not the expert here, just a XW6048 owner.
I'm pretty sure I've never heard or read "dynamically settable output current limit" anywhere. Is this regarding the AC output, or the DC charging output to the batteries ?

Anyway, the AC 120 hz ripple has to be supplied by the batteries. I also suspect that by adding a bank of Low-Z capacitors right at the inverter input terminals, you could relive a whole lot of that stress with caps, and get by with a bank of 8, 6V 200A golf cart batteries. You still have to have the batteries, for surges/outages AND you will need a load bank, because I'm sure the GTI will get glitched several times a day with out of spec voltage dips / peaks from the Grid, and you will have to dump your 5KW somewhere for the 5 minutes while the inverter resets. A couple of water heaters would do the trick I guess.


Candidate caps (I don't know if anyone has actually done this or not. (I'm considering it for a NiFe battery bank, to get the ripple off the batteries)
Panasonic
http://www.panasonic.com/industrial/com ... ES_DNE.pdf
http://search.digikey.com/us/en/product ... -ND/411629
or http://www.panasonic.com/industrial/inc ... -10-05.pdf

So, this is what I've been thinking about the last couple weeks, using a bank of caps (high ripple qualified) to reduce demand on the batteries.

I'd be interested to hear what SG has to say about the dynamically settable output power limit. It was explained to me like this: one can change the "Max Sell Amps" setting through the Xanbus network even while the unit is operating. So, if you're measuring the input current and output current of the GTI, you should be able to track well enough to maintain a relatively constant input current from the DC power system. That's really enough control for us. We would need some feedback for our own central controller, but we don't need split-second control here, just on the order of many seconds, which should be doable.

I've been thinking about capacitors at the input as well. This is what we may end up doing to cut down on the battery needs. I thought about super-caps, but I think that's more complicated and expensive than is justified.

The cutting out for 5 minutes may be alright. The requirements of the project are flexible, but the idea is to have near-continuous if not continuous operation. I'll have to consider some back-up water heaters though.

EDIT #1: BTW, concerning the caps again, this is where my analysis might be useful for spec'ing an array which can handle the ripple current I calculated. In almost all capacitor datasheets there is a maximum ARMS rating at 85°C specified. Honestly, something in this class of capacitors is probably a better bet: http://media.digikey.com/pdf/Data%20She ... Series.pdf. Check out the 63V rated caps in that PDF. Many of them have well over 10 ARMS as their ripple rating. You would need many of these in parallel to get the right effect.

EDIT #2: I did some extra calculations which show me that during the charge / discharge periods of each inverter half-cycle, a total of Q = ~0.30C of charge is being added / removed to the battery bank. Assuming that an input capacitor bank is handling all of this, and knowing that C = Q / V (the worst-case V still being 44VDC in this case), then a total of ~6.9mF is necessary to guarantee that the bank can handle this charge / discharge action while forcing delta-Q = 0. If you take a look at that U36D series, you can see that meeting the ARMS requirements is much more important than meeting the minimum capacitance requirements (i.e. I could easily use 7 x 1mF caps, but I wouldn't be able to handle the ARMS currents).

A possible solution is a minimum of six of these: http://search.digikey.com/us/en/product ... ND/2095943. This would be a total ARMS capability of 60.6ARMS, but would turn out to be a 132mF bank, costing $71. The series impedance of the cap-bank would be ~2.3mOhms, for a worst-case ripple voltage of 133mVRMS and maybe ~8W of heating throughout. This is not something to which one would connect a battery bank directly without pre-charging (at least I wouldn't).

Regards,
Tele

Xanbus is open and documented, so an application programmer could certainly program the maximum amps value dynamically and what ever other logic you need, typical for custom work.

The ripple current is a real issue. You need 100ah battery capacity per kw that is being exported to the grid. don't skimp on this, at best the batteries will have a shortened life, worst is you have thermal run-a-way and explosions. This is a battery issue and nothing to do with the XW. In true offgrid, for a 6kw inverter you would probably have a 2000+ ah battery, so this only showes up when people are trying to have "backup" but cheap out, with disastrous results.

Battery's could be augmented by capacitors. Keep in mind the best electrolytic's on the market typically has a ripple rating of 4-5 amps rms. So you looking at a bank of close to 100 of these, not to mention its not a UL approved installation if your a DIY for this.

Yikes ! After crunching my numbers, I came up with a 700ah bank, and upped it to 800ah. The current bank (400ah) of used L-16s shows no appreciable heating while a 1KW pump is running for 3 hours.

But I'd think any assistance from caps (suitably rated to not fail in 2 years) would help - even if the don't deal with all the ripple, which may not be 6kw for 8 hours at a time. I guess the caps should have their own fuse or breaker, and be in their own enclosure.

--


solar:http://tinyurl.com/LMR-Solar
Powerfab top of pole PV mount | Listeroid 6/1 w/st5 gen head | XW6048 inverter/chgr | Iota 48V/15A charger | Morningstar 60A MPPT | 48V, 400A battery bank | 15, Evergreen ESA 205 fa3 "12V" PV array on pole | Midnight ePanel || VEC1093 12V Charger | Maha C401 aa/aaa Charger | SureSine | Sunsaver MPPT 15A | Grundfos 10 SO5-9 with 3 wire Franklin Electric motor (1/2hp 240V 1ph ) on a timer for 3 hr noontime run - Runs off PV !

the 2000ah load is for typical depth of discharge using the inverters full capabilities. I can see what I wrote could be confusing, it was to highlight that gridtie aside from the ripple issue can operate without large batteries, hence why this ripple concern shows up in these implementations and not typical off-grid systems with the same inverter

Thank you both for the input. I think I am going to have to do some experimentation, and I'll let you know what I find out through the measurements I take. We have a 5kW power-supply that we can use to power the GTI, and we'll start taking away battery capacity until we find one that works when coupled with an input capacitor bank.

It's true that any added capacitance at the input would not be a UL-rated bank, but that's a problem for our power system, not for the inverter itself. I wasn't imagining installing the capacitor bank directly across the batteries, but rather only at the input.

The comments about the capacitors is confusing. The U36D series from United Chemi-Con claim much greater RMS ratings than 4~5 ARMS @ 85°C, depending on which one you choose. Second, needing 100 of them would imply that an RMS rating of 400~500 ARMS is needed, whereas my calculations show about 1/5th of that under worst-case (i.e. 5kW from the bank alone). Is it the case that, in practice, the ARMS ratings don't work to their specifications?

****************************

If you guys are interested, my analysis spreadsheet is attached. I've included comments to explain what I'm doing. You can judge for yourself whether or not my calculations are proper. If one enters 0 into the cell I have highlighted blue (PI,PS), the calculations are performed for 5kW of power coming purely from the battery bank. The DC current demanded from the bank would be 114A, and the RMS current 98ARMS. I would think that a 200Ah battery bank would be able to handle this for a few minutes at least, which is all we need. The amount of charge removed at Vin = 44VDC is equivalent to completely discharging a 22mF capacitor (see Cmin).

I propose installing an input bank of the following unit capacitor: http://search.digikey.com/us/en/product ... ND/2095946. The specs of the capacitor have been entered into the spreadsheet. Using a bank of 7 of these, I could form a 700mF capacitor bank having an ARMS rating of 150ARMS (50% more than worst-case to be encountered), an aggregate ESR of 0.8mOhms, and costing only $260 + engineering costs (which would probably double the price to ~$500).

During normal operation (PI,PS = 5kW), the bank would be discharged no more than 1% with each half-cycle of the current waveform (thereafter being re-charged by the input power). A total of 2.5W would be dissipated in the bank, and the voltage ripple would be 50mVRMS.

I do realize these are all idealized calculations, but with proper engineering (with some active cooling even), I think this bank could take most of the heat off of the battery bank. During periods where the input power is < 5kW, the capacitor bank would be of such low impedance that I'm pretty sure the battery bank would see much less of the AC current than normal. I have to verify this with experimentation.

Regards,
Tele

[EDIT: file deleted]

Yes, those look like a good caps to try out ... also, when I mention watts to battery ah general rule of thumb, I'm in general terms talking about AC watts, not DC watts as your looking at in the spreadsheet.

Just a suggestion, make the CAP array be in it own enclosure and have it have a bleed resistor ( safety ) and a DC breaker for protection against a fault. Who knows, maybe it would be a good product to sell if you can get it UL listed.

I'm use to having to fit caps into containers like the housing for charge controllers and inverters, so the 4-5a rms figure is the smaller case sizes. If your going to hand wire a bank of them ( watch the resistance ) the few larger ones are a good choice

good luck!

I guess a "soft charge up" device (relay & resistor) would be needed too, or there would be arc welding at initial power up.

The way we've solved this in the past is to use copper bus-bar to connect all the caps in parallel. Then again, high-gauge, high-strand-count battery cable might be better to reduce parasitic inductance. Then again, then again... the caps will be so close to one another that it shouldn't matter that much.

We've also used a soft-charge circuit just like you mentioned. Definitely not a capacitor bank one wants to hook batteries up to directly.

I'll try to update the thread when I have data to share. It'll be a few months probably.

Regards,
Tele