100hZ ripple from grid tie, economics of capacitors

[fourtytwo]

Hi everybody

I am in the process of designing a 500W grid tie, these reflect 100hZ ripple back to the PV array.
Does anybody have a rule of thumb as to how much it's worth spending on capacitors either on the PV input or the DC-link to improve the efficiency loss in the PV from said ripple ?
For example 1000uF/250V are £8/pair from RS (I don't buy big caps from Ebay) and that's nearly us much as the main transformer!!
I just get the feeling it's lot's of money for little gain.........opinions and experiences welcome
Great forum

Roger

 

[ronsimpson]

I usually work hard to keep the RF from getting into the array. (small caps)

You may have a graph for your array. Voltage, Current, Power. There is a maximum power point. My thought is that if the cap is very small and the voltage moves greatly then it would be hard to hit the max power point. You would run above and then below the point.

Another way to think about it: If there was no capacitor, during the zero crossing point of the power line, no current will come from the array. Not good.

Wild guess: I think you want the current from the array to only have 10% ripple at 100hz. This will keep you around the good point. (20% might work to but at some point the power level will drop)

 

[fourtytwo]

Thanks for the reply 10% is a good stab in the dark!
In an effort to answer the question without buying expensive components that turn out to be not needed I did a series of LTspice simulations today, each one takes 2 hours to run one grid cycle! but this simulates the PV's all the way to the grid.

Based upon two 72 cell panels in series giving 500W I get the following 120hZ ripple at MPP for different values of DC-link capacitor (with cost)
220uF £2.80 6.6% PV power ripple
470uF £5.70 3.6% PV power ripple
940uF £11.40 2% PV power ripple

Of course we all know that models are imperfect but at least it gives me some ideas of scale, hope this assists anyone else pondering the same questions.

Insofar as HF ripple this is all dealt with by polypropylene caps of <10uF my boost stage being two phase to reduce ripple!

As an aside I have never seen this figure quoted in commercial GTI's yet it is effectively an efficiency loss, the average PV power being reduced by half the ripple value.

 

[tcmtech]

Given you have two 72 cell panels in series you have a nominal output voltage of ~72 volts so at a ~500 watt output that a nominal current of ~7 amps.

Given that, my basic estimation would be to use around 250 uF per amp or a ~1500 - 2000 uF capacitor bank although depending on line losses and other unknown factors like GTI switching device zero cross dead band and peak current handling characteristics I would say that optimal could be anywhere between 1000 - 7000 uF.

 

[fourtytwo]

Thank you so much for sharing your experience, I guess you are talking about the value were the capacitor to be connected across the PV output, in my case they are at the 350VDC DC-link level however the CV value or energy stored is pretty much the same as that suggested by yourself. My concern or opinion would be when the capacitor/s are connected directly across the PV output the MPPT algorithm would be unable to closely follow the dynamic behavior of the panels, here in the UK our "summers" have frequent cloud density changes that the MPPT must follow rapidly for best performance.

 

[ronsimpson]

dynamic behavior of the panels

If we like the idea that the panel voltage can change 5% to 10% in 1/2 cycle of the power line, then a very small cloud will last 10,000 1/2-cycles.
I think the voltage will go from one limit to the other in well under 1 second. I don't your cap will make any difference.

 

[fourtytwo]

Yup on consideration I make it half a second BUT your haven't seen the high altitude winds here sometimes I am not talking whole clouds the differences in shade from light to dark grey, we rarely get whole identifiable clouds these days just shades of grey lols maybe in Colorado you get blue skies still Also living here in the countryside I have a lot of wood pigeons overflying the panels causing rapid fluctuations guess I am grasping at straws now to support my argument haha

 

[tcmtech]

Thank you so much for sharing your experience, I guess you are talking about the value were the capacitor to be connected across the PV output, in my case they are at the 350VDC DC-link level however the CV value or energy stored is pretty much the same as that suggested by yourself. My concern or opinion would be when the capacitor/s are connected directly across the PV output the MPPT algorithm would be unable to closely follow the dynamic behavior of the panels, here in the UK our "summers" have frequent cloud density changes that the MPPT must follow rapidly for best performance.

I don't know anything about that design of GTI systems other than I find them unnecessarily overly complicated.

If you look around in this forums Renewable Energy sections you will see I have a number of threads dedicated to the design of GTI systems and none of them use the HV DC intermediate stage to work.

That's where my hands on knowledge comes from. My actual having designed and built GTI using from the ground up.

 

[fourtytwo]

I don't know anything about that design of GTI systems other than I find them unnecessarily overly complicated.

If you look around in this forums Renewable Energy sections you will see I have a number of threads dedicated to the design of GTI systems and none of them use the HV DC intermediate stage to work.

That's where my hands on knowledge comes from. My actual having designed and built GTI using from the ground up.

Absolutely agree there, I find much of the published material overly complicated IMOP mostly caused by an addiction to digital signal processing and it's attendant delay's that causes substantially more complexity to overcome them! My proposed system returns to analogue processing with a handful of op-amps, after all its a simple enough analogue control problem! I do however prefer small lightweight ferrite transformers. I all way's apply the KISS principle and rigorously cost my projects to ensure they are economic and will pay for themselves in a few years at most otherwise they just get dumped! Once again thankyou for sharing your experience

 

[Tony Stewart]

Here's another perspective.

500W/230Vac RMS = 2.17A AC best case 100% eff. Linear load equiv= 106Ω
500W/350Vdc = 1.43Adc best case, Zeq=245Ω min.

Assuming MPPT is tracking ok, a 10% drop in Vdc should only drop Pout by 5%
In order to maintain -10% ripple , impedance of cap must be ~ 10% of Zeq or 25Ω
Thus at 100Hz C=1/(2π100*25)= 64 μF

The fault in this logic is the actual crest factor of Cap current to inverter.
If peak/RMS=3 then use 3x 64uF and divide by efficiency.

 

[fourtytwo]

Hello Tony,

I really like that way of looking at it and would agree entirely with your elegantly simple way of approaching the problem. The only thing I would mention is the relationship between Vdc and Pout, particularly if you consider the steepness of the MPP slope on the right a small change in voltage can make a large change in power and of course this relationship is highly dependent upon insolation level. Hence I went all the complicated and laborious method of making a complete model in LTspice being the only free tool available. I seemed to end up with around 500uF so far

 

[Tony Stewart]

I wasn't worried about the high side of MPT efficiency as this would be from low demand.
Although for 6£ @470uF, that is very inexpensive if rated above Voc

I'd get two
https://www.ebay.ca/itm/2PCS-New-47...394919?hash=item3ac2729f27:g:omgAAOSwubRXIYY4

 

[fourtytwo]

That was the bit that was worrying me, how much output loss would the ripple cause. Another way I think I could have done it was plot the MPP curve at different insolations on a piece of graph paper then draw the ripple voltage on there. Anyway thank you everybody I have a much better understanding of this subject now and multiple tools to use! BTW Tony every penny counts for payback time!

p.s. I made an error originally, they are 60 cell panels, hard to count as the corners are not cut off and they have more then the usual buss wires.

 

[MrAl]

Hello there,

To add a little new information here...

I would think that this should be handled just like any other cost analysis with associated future projections.
Since you have an estimate of the gains and the costs, you should be able to answer the question, "Is it worth it", with little or no problem.

The way it goes is just like anything else, where you purchase a product and it gives you a gain, and that gain is measured in dollars per unit time. You know this so you can project the "break even" point in time, where in the future the new product (the cap) pays for itself.
Once you get past this point of course then it is all profit, provided there is no maintenance on the product that would have to be factored in also.

So if you gain 5 percent efficiency and that results in 5 dollars extra per month and the cap costs 10 dollars, then the cap pays for itself in 2 months.

That's the right way to rationalize the cost benefits. If you have a question on efficiency vs cap size, it is probably best to experiment. Try a reasonably sized cap and see what you get over a couple months, then size up again if it looks promising.

I was speaking in terms of dollars (USD) but any type of currency would be basically the same.

 

[fourtytwo]

You are absolutely right Al, one of the problems I was having was working out the ripple voltage to efficiency reduction ratio and so was asking people who perhaps had experience of GTI's to give me some clues. The great thing about this community is that you can learn as you go along, so whereas at the beginning I was unsure of things like the method of my calculations I now know they not only have some validity but I have some other methods of approaching the problem thanks to the feedback here I am heavily into renewables and energy conservation in general at the moment but having designed products all my working life on a cost benefit basis I apply the same principle to all my home projects hence I do not throw money at a problem! Thank you so much for your lucid contribution

 

[tcmtech]

If you ever get around to experimenting with this I would like to see what you find for real-world efficiency gains based on your observations.

 

[fourtytwo]

Hi tcm, always happy to publish results, I have a blog here on that system. Might be a while yet as I am hardly making any surplus from water heating so probably wont build the GTI until I install some more panels ready for next summer and that will require management approval lols