Sanity Checking + Advice on SolarEdge

[aleenoor]

Thank you Sensij for clarifying that. I do actually have full access to my portal and I've played around just a tiny bit with the charts and reports features. I didn't realize though that I could go back in time and get historic data points that aren't available in playback anymore. That should be fun to play with, thank you! Can the API pull down those same historic data points, or are you limited to exporting them from the charts interface? I wouldn't mind doing a bulk data pull once a month, etc., to dump all that historic data to Excel...

Fun side story, I was the first person ever to ask my installer for increased portal access so that I could run charts and reports, and also the first to ask them for my API key. My PM wasn't really sure what the API key was, so she asked their IT department about it, and they got a little freaked out like "what is this customer going to be able to do with that key if we give it to him?". In the end, it was the electrician who had done the install and talked with me on-site that calmed everyone down and essentially said "whatever this customer wants just give it to him, he's an engineer"

Hi Marc,
Would you be able to share/email raw wattage data for your array on from a couple of 'good days'. I want to see what % age of time (and cumulative power) are modules producing over 90% of STC. That should give me an idea of how much I should be clipping in the best production months.

Thanks

[Magius]

No, optimizer data are not available through the API.

Thanks sensij, that's too bad. I was planning on playing around with the API in Python at some point, but hadn't actually gone to look and see what it could do yet. It seems like there's not much I could do with it, different from what PVOutput already does, so I may not even bother.

[Magius]

Aleenor,
I grabbed some data that should help you. There are three charts. The first is the raw wattage produced by two of my best panels on 3/14, which just happened to be a day that I was paying attention last month and noticed exceptionally high instantaneous power readings. These may or may not be the highest numbers recorded that month, but they really stood out at the time. Note that this was during an abnormal cold front (50s F) so everything was running extremely efficient. In addition, it was a cloudy day, so the peaks are due to cloud reflection. All that said, you can see that instantaneous numbers over 360W were recorded multiple times. 3-14.PNG


The second picture is of the following two days, 3/15 - 3/16 for the same two panels. Again, it was during the cold front, but the difference is these were very clear days. Nice smooth production curves all day long, this time just barely making it to the STC rating of 320W for a few minutes. 3-15_16.PNG


The third picture shows two similar days from about a week ago, 4/8 - 4/9, for the same two panels. The weather was significantly warmer, but similarly clear and sunny. As you can see, the panels just barely make it to 300W the first day, and don't quite get there the second day. This is what I consider "normal" behavior, when not pushed higher by cold weather or scattered clouds. 4-7_9.PNG


So your question was, "how often will the modules produce 90% or more of their STC?" (ie: 288W or more). I took one of the two days in that third photo and blew it up, to mark out what portion of the day was above your 90% mark. image_9088.png


To a loose approximation, it looks like ~90 minutes. Of course, we're not even into the summer yet, where temperatures will be 10-15 degrees hotter, voltages will drop, and that time window will get shorter. I'm not sure what drove you to choose 90% of STC as your metric, but with that yardstick, I'd wager this is loosely what you have to look forward to.

[J.P.M.]

Aleenor,
I grabbed some data that should help you. There are three charts. The first is the raw wattage produced by two of my best panels on 3/14, which just happened to be a day that I was paying attention last month and noticed exceptionally high instantaneous power readings. These may or may not be the highest numbers recorded that month, but they really stood out at the time. Note that this was during an abnormal cold front (50s F) so everything was running extremely efficient. In addition, it was a cloudy day, so the peaks are due to cloud reflection. All that said, you can see that instantaneous numbers over 360W were recorded multiple times. 3-14.PNG


The second picture is of the following two days, 3/15 - 3/16 for the same two panels. Again, it was during the cold front, but the difference is these were very clear days. Nice smooth production curves all day long, this time just barely making it to the STC rating of 320W for a few minutes. 3-15_16.PNG


The third picture shows two similar days from about a week ago, 4/8 - 4/9, for the same two panels. The weather was significantly warmer, but similarly clear and sunny. As you can see, the panels just barely make it to 300W the first day, and don't quite get there the second day. This is what I consider "normal" behavior, when not pushed higher by cold weather or scattered clouds. 4-7_9.PNG


So your question was, "how often will the modules produce 90% or more of their STC?" (ie: 288W or more). I took one of the two days in that third photo and blew it up, to mark out what portion of the day was above your 90% mark. image_9088.png


To a loose approximation, it looks like ~90 minutes. Of course, we're not even into the summer yet, where temperatures will be 10-15 degrees hotter, voltages will drop, and that time window will get shorter. I'm not sure what drove you to choose 90% of STC as your metric, but with that yardstick, I'd wager this is loosely what you have to look forward to.

FWIW, IMO, nice examples. A picture is truly worth a thousand words. Thank you.

[Magius]

FWIW, IMO, nice examples. A picture is truly worth a thousand words. Thank you.

Thank you! I appreciate that.

I also neglected to include an additional few thoughts on what the pictures really amount to when it's said and done. I thought that by doing some really really rough math it could illustrate the (in)significance of that little hump over the 90% line. I'm still not 100% sure what the 90% line was supposed to represent, but that's what I was asked to use so I'll stick with it

So I figured, lets assume that hump was a triangle. We know the height is 12 (300-288 W) and we'll round the base up to 2 (for 2 hours). Area of that triangle would then be 12 Wh. This should be reasonably more area than the actual shape which is more jagged and does not extend for a full two hours. So that's <12 Wh lost per south-facing panel, per (clear, spring) day. If we further extrapolate that to assume the same loss happens every single day of the year (which it clearly does not), and apply 10c/kWh as the cost of electricity, we come out with every south-facing panel losing <$0.44 worth of production per year. If you're going to install 30 panels, that'd be ~$13/year in lost production across the whole system. In reality of course, you won't see this behavior most of the year, since it's too hot in the summer and there's not enough irradiance in the fall and winter. Actual lost production would more likely be less than 1/3 of that bogus number (change the assumption to it clips this way every day for 4 months of the year, which is still too high). That would be ~$4.00/year lost across a whole 30-panel south-facing system. Of course, depending on where you set that clipping threshold, say at 80% instead of 90%, the numbers would increase fast, so I'm not advocating throwing power away willy-nilly!

Hopefully I didn't botch the units somewhere and make a fool of myself with bad math, but long story short it's not worth chasing a couple of dollars over to these clipped peaks for short periods. Higher STC panels will more than make up the difference with increased production on overcast days and in the mornings/evenings. Just trying to keep things in perspective, as I'll be the first to admit I over-obsessed about this same phenomenon until I got a little more familiar with the quirks.

[aleenoor]

Marc, really appreciate it. Basically, the 90% was a ball part. My Inverters (10kW) are about 85% of the STC capacity of my panels (40x295 = 11.8kW DC) accounting for losses at peak production time (conductors ~1% , inverter efficiency 96.5~). So I should start clipping anytime my panels are producing over 88% of STC. I wanted to gauge how much time will they be doing that on the best of days.

I am trying to decide if I should switch to 2x 4Kw inverters from 2x5kW. Or shell out another $2000 for Solaredge 11.2kW + P300.

[Magius]

I am trying to decide if I should switch to 2x 4Kw inverters from 2x5kW..

I see now where you're coming from and where you got the 90% figure from. For what it's worth, I'm no expert by any means, but I would absolutely not recommend running 11.8kW worth of all south-facing modules into two 4kW inverters. That's 47.5% oversizing, and you'll be throwing away a ton of power during the peak of the day. If I applied that kind of oversizing on my 18.24kW system, it would equate to having only 12.36kW worth of inverters. Even though my system is spread across 3 azimuths, with only 1 in 5 panels facing south, I've peaked at ~15.8kW, and I sure wouldn't want to throw away 3.5kW due to clipping, even if it was only for short periods. Not to mention with your all-south setup the ratio of waste would be higher than mine...

So definitely stick with the 2x5kW inverters at a minimum. With those you're at 18% oversized, similar to my system which is 20% oversized, but unlike mine you expect to point everything due south. Doing a little extrapolation, if my south-facing 320W modules top out at 300W on an average day, then your 295W modules should top out around 277W. x40 that's a hair over 11kW of production, so with two 5kW inverters you're looking at potentially clipping ~1 kW of power at the peak of the day, or ~25W per panel. That's essentially double the 12W/panel that I was working with in the example above with the "triangle" math, plus you have 40 panels vs. the 30 I assumed. So a "high" estimate for the power lost to clipping might be (2x$4)*1.33 or <$11/year.

I'm obviously a fan of SE, but spending $2000 more to prevent losing <$11/year wouldn't make much sense if you're just talking finances. Over 20 years eating the clipping is still an order of magnitude cheaper Now if you have shading issues that the P320's might help mitigate, or if you really want per-panel monitoring, or rapid shutdown, then sure, the $2k might provide other tangible value. But I can't say I'd do it in your case solely to prevent inverter clipping.

[aleenoor]

Marc, really appreciate it. Basically, the 90% was a ball part. My Inverters (10kW) are about 85% of the STC capacity of my panels (40x295 = 11.8kW DC) accounting for losses at peak production time (conductors ~1% , inverter efficiency 96.5~). So I should start clipping anytime my panels are producing over 88% of STC. I wanted to gauge how much time will they be doing that on the best of days.

I am trying to decide if I should switch to 2x 4Kw inverters from 2x5kW. Or shell out another $2000 for Solaredge 11.2kW + P300.

So I took your 2nd graph where you Panles cleanly hit thier STC of 320W. Per my pixel based analysis, they were at 320W or higher for almost 4 hours. Since th graph is more a parabola than triangle, I calculated area under it as 4hrs X 40 W X 0.6 (vs 0.5 for a triangle). = 96Wh

Also, I drew the line at 280W which is 87.5% of STC. >> After Losses, my inverter is 87.5% of my DC STC power.

Since my Panels only produce 295 / 320 = 92% of your panels my per panel loss looks like 0.92 X 96Wh = 88.5Wh.

My Array loss should be 40X88.5Wh = 3.5kWh.

Lets assume I clip about 75 Days out of the year ( too optimistic ?) my total clipping losses should be 262 kWh annually.

That is about $35 / year or about 1.5 ~2% of my annual production.

Bad math ???

[aleenoor]

I see now where you're coming from and where you got the 90% figure from. For what it's worth, I'm no expert by any means, but I would absolutely not recommend running 11.8kW worth of all south-facing modules into two 4kW inverters. That's 47.5% oversizing, and you'll be throwing away a ton of power during the peak of the day. If I applied that kind of oversizing on my 18.24kW system, it would equate to having only 12.36kW worth of inverters. Even though my system is spread across 3 azimuths, with only 1 in 5 panels facing south, I've peaked at ~15.8kW, and I sure wouldn't want to throw away 3.5kW due to clipping, even if it was only for short periods. Not to mention with your all-south setup the ratio of waste would be higher than mine...

So definitely stick with the 2x5kW inverters at a minimum. With those you're at 18% oversized, similar to my system which is 20% oversized, but unlike mine you expect to point everything due south. Doing a little extrapolation, if my south-facing 320W modules top out at 300W on an average day, then your 295W modules should top out around 277W. x40 that's a hair over 11kW of production, so with two 5kW inverters you're looking at potentially clipping ~1 kW of power at the peak of the day, or ~25W per panel. That's essentially double the 12W/panel that I was working with in the example above with the "triangle" math, plus you have 40 panels vs. the 30 I assumed. So a "high" estimate for the power lost to clipping might be (2x$4)*1.33 or <$11/year.

I'm obviously a fan of SE, but spending $2000 more to prevent losing <$11/year wouldn't make much sense if you're just talking finances. Over 20 years eating the clipping is still an order of magnitude cheaper Now if you have shading issues that the P320's might help mitigate, or if you really want per-panel monitoring, or rapid shutdown, then sure, the $2k might provide other tangible value. But I can't say I'd do it in your case solely to prevent inverter clipping.

Hi Marc,
read this after I had posted. Yeah both maths make sense. Considering between 3x 4k or 2x 5k inverters. 3x 4k will be $500 more.

[sensij]

So I took your 2nd graph where you Panles cleanly hit thier STC of 320W. Per my pixel based analysis, they were at 320W or higher for almost 4 hours. Since th graph is more a parabola than triangle, I calculated area under it as 4hrs X 40 W X 0.6 (vs 0.5 for a triangle). = 96Wh

Also, I drew the line at 280W which is 87.5% of STC. >> After Losses, my inverter is 87.5% of my DC STC power.

Since my Panels only produce 295 / 320 = 92% of your panels my per panel loss looks like 0.92 X 96Wh = 88.5Wh.

My Array loss should be 40X88.5Wh = 3.5kWh.

Lets assume I clip about 75 Days out of the year ( too optimistic ?) my total clipping losses should be 262 kWh annually.

That is about $35 / year or about 1.5 ~2% of my annual production.

Bad math ???

This is absurd. Perfectly capable modeling tools exist that could help you answer this question (see PVWatts, or SAM, if you want to take it to the next level). Basing your design around the clipping you calculate on an abnormally cold day is spring is not going to result in a very cost-effective decision.

[aleenoor]

This is absurd. Perfectly capable modeling tools exist that could help you answer this question (see PVWatts, or SAM, if you want to take it to the next level). Basing your design around the clipping you calculate on an abnormally cold day is spring is not going to result in a very cost-effective decision.

Hi Sensij,
I am not aware of SAM but in PV watts, even choosing a weather station 5 miles apart result in big variance. Also, I realize this will not be a consistent occurrence hence accounting for about 75 days..

[Magius]

Yeah both maths make sense. Considering between 3x 4k or 2x 5k inverters. 3x 4k will be $500 more.

In your earlier post you said you were looking at switching to two 4kW inverters from 5kW. Now that you say three it makes a lot more sense
Even with your version of the math, which I don't think is any more or less "bad" than mine, you came out with spending $500 up front to save $35/year. That's ~15 years to break even, assuming your math is accurate, which I think we all agree it was not meant to be. At least on my part this was an exercise to show magnitudes, and clearly the magnitudes do not support buying more inverter just to avoid clipping. This was what I was trying to demonstrate all along.

This is absurd. Basing your design around the clipping you calculate on an abnormally cold day is spring is not going to result in a very cost-effective decision.

I agree completely. I admitted up front that the math I was going to show was absurd, but the intent was to provide an example in some type of quantitative way that none of this short-term clipping is worth worrying about. I further agree that using that one abnormally cold spring day where I peaked at 320W is even more counter productive, which is why I used a "typical" spring day with a 300W peak in my calculations. Even so, doing it aleenoor's way with the unicorn day, his math shows (IMHO) that it isn't worth upgrading his inverter, so I think the point has essentially been beaten to death

[J.P.M.]

Hi Sensij,
I am not aware of SAM but in PV watts, even choosing a weather station 5 miles apart result in big variance. Also, I realize this will not be a consistent occurrence hence accounting for about 75 days..

FWIW, I pretty much come down w/ Magius and Sensij on the futility of either worrying a whole lot about clipping and/or spending a whole lot of $$ on inverter upsizing for a mostly unknown (with any precision) but likely slight loss of performance. Hell, unless you hose the array 1X/month or so, you'll probably lose 2-3 % in annual performance from a dirty array and never notice it. Out of sight, out of mind.

On location variance, both from m odels and observation: Expect it. And also keep in mind, or read up on, the limits of models. Start w/ the PVWatts help/info screens.

Given any of the model's limitations, the weather's variability and instrument limitations, I'm pretty sure you can't measure, much les small amounts of clipping with the necessary precision to make estimates of lost performance with the necessary accuracy or precision for economic or costing decision making purposes.

It may be an interesting academic pursuit, but results may not justify the expense if you're looking for answers hard enough to justify spending extra $$ to get rid of clipping.

1.) On either SAM or PVWatts, the irradiance data comes from the same source(s).
2.) That data is, for the most part, modeled itself, that is, mostly but not entirely modeled from some ground observations at selected sites and then extrapolated to other sites along with such other weather data that may exist at those minor sites. See the TMY manual (SAM and PVWatts use TMY data or other data from "SolarAnywhere") for particulars on modeled data, the extent of modeling and modeling limitations on accuracy of estimates.
3.) Any solar model's output for any site is no better than an estimate of long term average performance. Using any model's output and expecting it to lead to reliable estimates of long term performance, even if smoothed out or using many (75 as you say) discrete points and summing the results to predict daily or hourly performance or clipping will lead to problems.

Think about this: Unless there is a significant mismatch between inverter and array capabilities, clipping is likely to be of the order of a few % or so of max. output, for short periods of time, for a few days/yr. -- Peanuts. On any "clear" day, irradiance input over the course of an hour of several hours is likely to vary by (also) a few% or so from what any irradiance model might estimate. That's what I learned over 40+ yrs. of textbook scouring and then confirmed by using professional grade instruments (Eppley) and the Davis weather station on my roof that measures Global Horizontal Irradiance and records that and other data at 1 min. increments. The models are good, but best for long term average estimates. That's partly due to the nature of the models, but mostly because the atmosphere is quite variable.

Bottom line: Worry about clipping if you choose, but odds are it ain't a big deal, and, in any case, difficult to impossible to long term quantify as to any economic penalty, will decrease over time as array performance deteriorates, and until you, I and the rest of the world get better at measuring it, have performance penalties that are likely small enough to get lost in instrument variability and signal noise.

Take what you want of the above. Scrap the rest.

[ButchDeal]

Aurora design tool gives us a % inverter clipping. This is with actual modeling and time based so you can fit a lot more modules on an inverter with east/west vs south only, and it includes time based shadows.
We do our designs such that that is limited to less than 1% but that is just our internal policy. It costs us little to move up from one solaredge inverter to the next. We occasionally have exceptions to the rule, usually for MSP limitations but we keep them small as well.

We have seen many competitor designs that include considerable clipping so I agree that in this case the clipping is nothing to worry about, I disagree that customers should not worry about a potential system by their installer having clipping.

[J.P.M.]

Aurora design tool gives us a % inverter clipping. This is with actual modeling and time based so you can fit a lot more modules on an inverter with east/west vs south only, and it includes time based shadows.
We do our designs such that that is limited to less than 1% but that is just our internal policy. It costs us little to move up from one solaredge inverter to the next. We occasionally have exceptions to the rule, usually for MSP limitations but we keep them small as well.

We have seen many competitor designs that include considerable clipping so I agree that in this case the clipping is nothing to worry about, I disagree that customers should not worry about a potential system by their installer having clipping.

Is that 1% limitation on max. instantaneous output or 1% limitation on annual total output, or something else ?

I'd agree that if an inverter will clip to the point that it reduces the nominally expected annual output by, say, 1%, that may be a worthy figure provided that the cost of the lost production represents a significant proportion (say 10% or so ??) of the incremental cost of upsizing an inverter. A 1% annual clipping for me might amount to, say, 90-100 kWh/yr. At ~ $0.26/kWh = $26/yr. it may be worth it to upsize the inverter - but a bit off topic - provided I'm not an excess generator.

If I am an excess generator, downsizing an array by removing panels (at the design stage), and (possibly) keeping the same inverter becomes a more cost effective solution than increasing inverter size. Other things such as future expected needs may well dictate the need for initial excess generation however.

The usual goal is the lowest long term cost of providing electricity through a combination of a properly sized and designed PV system and POCO supplied electricity. I'd suggest that some clipping may be in line with system design and cost goals.