Oversizing inverters

[bcroe]

Considering a decade aging, converter losses, and DC wiring
losses (loops up to 400 feet), the effective DC:AC is around 2.1:1.
To my surprise, measurements in 2016 indicated I hardly needed
any south facing panels for best performance. That upgrade has
not been done, but there is still room for getting more energy from
the 15KW inverter plant.

The peak power applied is probably more like a 1.2:1 at any time,
more measurements could check. Good luck getting the inverter
mfr to approve this. Working here for 10 years. Bruce Roe

[nerdralph]

I think it's a bit of a red herring, as the best ratio is highly variable and site specific and is one of many factors in LCoE. However, if you can't do the LCoE calculations, 1.25:1 will usually be within 5% of the optimal ratio.

[nerdralph]

Here's a Solis 1P-5K-4G-US inverter that I'm intentionally stress testing with a 1.5:1 DC:AC ratio. The clipping is obvious.
Time, W AC, W DC, efficiency
2023-04-28T11:42:13,5080,5373,94.5%
2023-04-28T11:47:24,5080,5368,94.6%
2023-04-28T11:52:36,5080,5386,94.3%

And here's a Solis 1P-7K-4G-US inverter with a 1.05:1 DC:AC ratio.
Time, W AC, W DC, efficiency
2023-04-28T11:42:18,5740,5974,96.1%
2023-04-28T11:47:18,5840,6045,96.6%
2023-04-28T11:52:29,5910,6170,95.8%

[bcroe]

And here's a Solis 1P-7K-4G-US inverter with a 1.05:1 DC:AC ratio.
Time, W AC, W DC, efficiency

2023-04-28T11:52:29,5910,6170,95.8%

Keep in mind, with an efficiency of 95.8%, the input DC level must
reach 1/.958 = 1.044 of output to cover loses, before clipping starts.
Bruce Roe

[Ampster]

.......The clipping is obvious.
......

Yes the clipping will be obvious. What is not onvious, but more important long term is what is the difference in production over a typical year. You can get a crude estimate from PV Watts.
A three minute group of numbers is meaningless for making a decision about a long term investment like a solar system. Statistically the sample size is to small.

[nerdralph]

Keep in mind, with an efficiency of 95.8%, the input DC level must
reach 1/.958 = 1.044 of output to cover loses, before clipping starts.
Bruce Roe

Agreed. I'm adding a third string with 2.7 kW, and will flash it with the 1P8K firmware, so the ratio will end up around 1.26:1.

[J.P.M.]

Here's a Solis 1P-5K-4G-US inverter that I'm intentionally stress testing with a 1.5:1 DC:AC ratio. The clipping is obvious.
Time, W AC, W DC, efficiency
2023-04-28T11:42:13,5080,5373,94.5%
2023-04-28T11:47:24,5080,5368,94.6%
2023-04-28T11:52:36,5080,5386,94.3%

And here's a Solis 1P-7K-4G-US inverter with a 1.05:1 DC:AC ratio.
Time, W AC, W DC, efficiency
2023-04-28T11:42:18,5740,5974,96.1%
2023-04-28T11:47:18,5840,6045,96.6%
2023-04-28T11:52:29,5910,6170,95.8%

Pardon my ignorance. I'm confused.
Are the inverters driving equal arrays ?

[J.P.M.]

Yes the clipping will be obvious. What is not onvious, but more important long term is what is the difference in production over a typical year. You can get a crude estimate from PV Watts.
A three minute group of numbers is meaningless for making a decision about a long term investment like a solar system. Statistically the sample size is to small.

I agree that the sample size is inadequate but I'm not sure extrapolating long term output from an inadequately small sample was what Ralph had in mind.
Ralph ??

What I'm more sure of is that an inverter with a maximum output that's smaller than the array's maximum new and clean instantaneous output will clip some depending on ow much it's undersized as Ralph's post may be showing.
I'm also pretty sure that an inverter that's theoretically sized to just meet an array's maximum new and clean instantaneous output won't clip under any conditions approximating a quasi-steady input.
A properly running inverter that's larger than the array's maximum new and clean instantaneous output probably won't clip but will not as cost effective as a larger (oversized) inverter because it'll probably cost more than a smaller inverter. That may be a design consideration however depending on future expansion plans or other application dependent requirements.
I'm also sure most folks only care about clipping and things like DC/AC ratios only to the extent it affects their bottom line.

To that end I offer some idea of a method to put some analysis behind the guesses.

I appreciate, maybe more than most/ the importance of cost effective design and I understand the idea that using an inverter whose output is less than an array's maximum new and clean instantaneous output (less inverter inefficiency) will avoid some underutilization of an inverter's capacity. I also understand that increased utilization of an inverter's capacity comes at a price. That price takes the form of some clipping.
What to do ?
Seems like there may be a sweet spot inverter size there somewhere but how to find it ?
Well, one practical approximation tool is PVWatts. The DC/AC ratio can be useful as well as the hourly output option and those who are familiar with the PVWatts model will find it easy to get a reasonable 1st approximation of clipping kWh.

Using my array as an example: 5.232 STC kW with a 5 kW inverter (DC/AC ratio = 1.0464) dc/ac. actual 10yr. average annual output = 9,036 kWh/yr.
PVWatts modeled output = 9,039 kWh/yr. system loss parameter == 13.4 % including 3.5% shading loss in late afternoon.
If I reduce the inverter size to 4kW (with DC/AC ratio = 1.308): Modeled output = 8,997 kWh/yr.

I did 2 runs, using my 5.232 STC kW array.One with a my current 5 kW kW inverter and one with the same array but a with 4 kw iinverter.
Some differences in the two runs:
Annual output drops by 39 kWh/yr. for the 4 kW inverter vs. the 5 kW inverter.
As might be expected, no clipping occurs with the 5 kW inverter.
The 4 kW inverter sees any clipping a total of 273 hours/modeled year out of 4,353 hours per modeled year of inverter operation. The 5 kW inverter sees no clipping as might be expected (real world and observed reality is however that there is about 3 kWh/yr. of clipping from cloud reflections that usually last < 30sec. and bump the inverter output into clipping. That happens maybe 30-50 times/yr., always +/- couple of hors around solar noon).

I rooted around on the net earlier today and estimate that the cost differential of string inverter in the 4 to 8 kW is ~ $0.35 meaning a 1kw step up in inverter (and assuming it can be installed for the same labor rate as a smaller inverter) will cost about $350 or so.
The current average cost around here for a residential T.O.U. kWh used in the afternoon when clipping might most commonly occur is something like $0.47/kWh.

Using the moron method of payback: $350/(39 kWh/yr.*$0.47/kWh) = ~19.1 years. Not looking too good but a more thorough cost analysis using LOCE or other methods would probably produce a more cost-effective result. How much more would depend a lot on the assumptions/estimates that were made with respect to the time value of money.

The above is for illustrative purposes only and may be useful as a go-by to put some additional method behind the inverter sizing/oversizing discussion.

[nerdralph]

I agree that the sample size is inadequate but I'm not sure extrapolating long term output from an inadequately small sample was what Ralph had in mind.
Ralph ??

My intention was to provide that 10-minute sample of data to show two extremes of oversizing ratios. And you are correct that I did not suggest or imply the data should be extrapolated over the whole year.

However I think anyone with significant knowledge of the seasonal patterns of PV system output can likely extrapolate a fair bit from that limited amount of data. For example you could extrapolate the temperature coefficient of the panels and conclude that the 1.05:1 system clips on cold (around 0C), clear, sunny days in March and April. And you could conclude that there will be no clipping throughout the summer since there is no clipping on a mild and sunny day in April.

[davidcheok]

[QUOTE=J.P.M.;n439262]

.....a more thorough cost analysis using LOCE or other methods would probably produce a more cost-effective result. How much more would depend a lot on the assumptions/estimates that were made with respect to the time value of money./QUOTE]

This. It is the same reason why energy producers use this method for system design for better LCOE/ROI. There are also other factors that solar calculators do not take into account like drop in output of the array over the lifetime of the system. This coupled with the higher temperatures of panels that varies depending on solar intensity often reduce real-world output.

[J.P.M.]

My intention was to provide that 10-minute sample of data to show two extremes of oversizing ratios. And you are correct that I did not suggest or imply the data should be extrapolated over the whole year.

However I think anyone with significant knowledge of the seasonal patterns of PV system output can likely extrapolate a fair bit from that limited amount of data. For example you could extrapolate the temperature coefficient of the panels and conclude that the 1.05:1 system clips on cold (around 0C), clear, sunny days in March and April. And you could conclude that there will be no clipping throughout the summer since there is no clipping on a mild and sunny day in April.

Thank you for the reply.

FWIW, I agree with your thoughts/opinions as presented above and their general validity. I wouldn't bet the farm on it holding true for all situations as it's not a rigorous analysis - but it's not being represented as one and that's fine. This is not a peer reviewed journal.

[J.P.M.]

[QUOTE=davidcheok;n439268]

.....a more thorough cost analysis using LOCE or other methods would probably produce a more cost-effective result. How much more would depend a lot on the assumptions/estimates that were made with respect to the time value of money./QUOTE]

This. It is the same reason why energy producers use this method for system design for better LCOE/ROI. There are also other factors that solar calculators do not take into account like drop in output of the array over the lifetime of the system. This coupled with the higher temperatures of panels that varies depending on solar intensity often reduce real-world output.

What is "This" ?

[Ampster]

............I offer some idea of a method to put some analysis behind the guesses... ...,.....

I use the same simple analysis and agree that comparison of annual output of alternate designs is the best way to find a cost effective DC to AC ratio assuming all other design parameters are the same.
in summary, the important numbers using a 5.232 array are:
4 kW inverter estimated annual production 8,997 kWh
5 kW inverter estimated annual production 9,036 kWh
39 kWh extra production from upsizing the inverter.
If someone were to focus on clipping they might be fooled by the 273 hours of clipping and might spend ,$350 essentially to save $18 per year or as you point out, close to a twenty year payback..

[DanS26]

Over the years we have seen many of these "clipping" threads. I think it is this "waste not" attitude of people not familiar with PV production. Inverters do not shut down when clipping, they just throttle back.

It is amazing to me that we have to convince people to oversize the array rather than oversize the inverter. It is a counter intuitive subject that takes a lot of explaining to have people to "step out of the box".

[Ampster]

Over the years we have seen many of these "clipping" threads. I think it is this "waste not" attitude of people not familiar with PV production. Inverters do not shut down when clipping, they just throttle back.

It is amazing to me that we have to convince people to oversize the array rather than oversize the inverter. It is a counter intuitive subject that takes a lot of explaining to have people to "step out of the box".

And yet clipping is still mentioned even in a response that primarily is about annual production. It is so much simpler to just look at annual production estimates when trying to decide between two alternaives. Clipping is a distraction the prevents many people from seeing the important issue which is production.