Putting a battery on power supply output to help with load transient performance

Hi
We are doing a 12kW output power supply with....
Vin = 700VDC
Cable resistance = 7.08 Ohms
Vout = 48VDC
Its an SMPS made with DCDC modules.

The load will have transients which are likely to be from no_load to 130% overload and vice versa.
Overload is treated with an output current clamp, but this takes some time to act (approx. 150ms).

In order to service these transients, (without the power supply tripping out) we are thinking of having a lithium battery simply connected to the output. (We need one there anyway to act as a power fail power source).

Is this concept of using a battery on a power supply output a common way of meeting heavy load transients as described?

There won’t be any interfacing circuitry (other than an initial inrush cct so that the battery doesn’t inrush to the power supply output caps upon connection) . The battery will start off being fully charged, so will only take a “trickle” charge from the power supply.
We are thinking of picking a ~47V battery so that it settles to 48V and just gets trickle charged as the power supply supplies the load.

Its a technique commonly used but you have to be careful not to switch the PSU on while the battery is discharged. A trick that might work is to wire a (big!) schottky diode and a resistor in parallel and wire them in series with the battery. The resistor limits the current in both directions but the diode conducts if the load drops the voltage more than Vf from battery voltage.

Brian.

Because batteries are chemical reaction based, they may
have a leisurely roll-on of current (considered at load-step
timescales). They are not just oversized capacitors (whose
response is electrostatics, simple carrier movement and not
ions in electrolyte).

If you want to go 0 - 250A in a millisecond, can any battery
do that? I'm suspicious that it's "no", as I have seen some
folks combining supercapacitors with batteries to handle
load transients -on the battery-.

The right sized Li-Fe-PO4 can easily be used as a transient OV suppressor, 50A-hr is good for up to 100 A, and so on ( prismatic type )

- - - Updated - - -

assuming they are not fully charged to start with ...

Is this concept of using a battery on a power supply output a common way of meeting heavy load transients as described?

Click to expand...

Should not be difficult.

Looking into the power supply, battery should not be a heavy load; the battery should be ideally floating all the time.

With a 12kW load on the far side, and with overloads, you need to oversize.

In a steady state, there should be no problems.

Should be ok, as long as complies with the following rules:
1. Batteries cranking capacity (starting or short term pulse, SAE) to be bigger than than your 130% short-time load requirement for transient time. In this case, 12kW at 48V, makes 250A, 130% will be 325A. Lead or gel car batteries have something like 550 to 800 A cranking current (55-75Ah). Did not have experience with Ni-Cd this size, but any producer will provide data for them.
2. Voltage of these batteries must be same as output voltage, means 48Vdc.
That makes 4x12V, meaning 4 pcs. batteries. For other voltage, do the math yourself.
Lower capacity of batteries will be useless, higher will be expensive.
Of course, must be done maintenance of them on a regular basis.

Using a battery for load leveling is fine, but you'd be foolish to think you can just treat the battery like a big capacitor by connecting it with diodes and resistors. You need a proper charge management system, especially if it's lithium based.

As the intent is load-step flattening, I'd begin with some
switched-load testing and see what the battery output
voltage sag transient might be. You may still need some
fat caps to soak up the leading edge until the chemistry
comes on line. This all would depend on the spec load
max dI/dt (plus any margining for conditions, aging,
repetitive events that might pull down charge faster
than it's added, etc.). This latter especially if max charge
by the "charge controller" if any, is held much less than
max discharge (like NiCd wanted C/10 charge I recall?).

You may still need some fat caps to soak up the leading edge until the chemistry comes on line.

Click to expand...

According to my experience as well as battery theory and commonly used battery equivalent circuits, there's no delayed response. Batteries can be modelled by RC ladder circuits and a small series inductance representing the current path geometry. Bypass capacitors are useful if the load has considerable current ripple, otherwise not required.

As mentioned by Easy peasy, Li-Fe-PO4 batteries can be probably used in simple parallel connection with constant voltage trickle charge. Li-Ion would need dedicated BMS and control of the primary power source according to battery SoC.

Thanks, So to summarise, and to be more precise, We are doing a 12Kw power supply with vout = 47.18V at max load.
Though At 10% load, the vout is 49.45V.

That is, the vout rises linearly as the load current falls. (this is the modus operandi of the DCDC modules and facilitates paralleling).

The problem is that there will be no_load to full_load transients, and so we need a battery to be on the output to help handle these. (the battery is also needed for power supply failure management).

What nominal voltage of lithium battery would you pick for this?

I am thinking that we can obviously not pick a battery voltage that’s less than 49.45V, because it would get worn out by the constant trickle charging?
Therefore, we must pick a battery of say 50V. However, this means that it will get discharged down to 47.18V (as most of the time the product will be on about maximum power) …..which means it will not last long as batteries need to be returned regularly to their nominal voltage in order to last a decent lifetime?
(It must be a lithium battery as small size is essential.)

a good quality Li.Fe.PO4 prismatic cell is 3.0 volts nominal and 3.3V abs max, so 16 cells give 48V nom & 52.8 abs max ...

Pretty sure LiFePO4 is 3.2 nominal and 3.6 max, regardless of packaging.

The battery lifetime will be limited. How much energy do you need for the power fail feature?
Maybe supercaps can solve your problem. They will have a much longer lifetime, and they are even better than batteries to instantly deliver a lot of current.
You need some balancing circuit for such a series connection of supercaps, but that is very common now.
You maybe need some charging logic if you want to start with discharged supercaps.

Edit:
I now see that small size is important, and that will be a problem if you want a lot of energy for the power fail situation.
If you only need the instant current, maybe supercaps can compete with lithium batteries.

they are even better than batteries to instantly deliver a lot of current.

Click to expand...

only the very best super caps, i.e. expensive, have the necessary low internal resistance to

.. instantly deliver a lot of current.

Click to expand...

try reading the data sheets and you will learn this ...

The problem is that our DCDC modules (as discussed above) reduce their vout as their output current increases. (and obviously vice versa)
As such, if the load is maximum for a long period, then the Lithium battery will get discharged down to 47.18V…subsequently if the load then gets lighter for a long period, then our power supply vout will rise, and potentially put too much current into the (now fairly discharged) battery….and we are not allowed to put more than 90A into the battery.
It seems the only way round this is…

1…Use a battery management system (awkward because the load is not ours and it doesn’t have a BMS, and we don’t have permission to add one to it)

2….Change the power supply to one with a constant voltage output, and make this Vout equal to the battery’s fully charged voltage….then we have to hope that the customer will always have the battery fully charged whenever connecting up to our power supply (otherwise we could end up overcurrenting it with >90A).
Being constant Vout would unfortunately get rid of the nice “denigrating_vout_paralleling_assistance” feature however, we could add current clamps to each DCDC module output and just ensure no DCDC module gets too hot like that…it would mean poor lighter load sharing between the DCDC nodules, , but who cares as long as no DCDC module overheats.

As you know, there are other fixes, but we are on a shoestring budget.