ground a solar farm

[PVfarmer]

Would 2 sets of strings each feeding its own inverter but part of the same PV system be two arrays?

I really don't think so.
A "set of strings" to me could mean anything from one to...not sure. 12 strings into one combiner?

You could have two strings, one per inverter, or 12 strings, 6 per inverter- that's two inverters and ONE array to me.
"PV system" to me means everything connected to one AC combiner.

Say you have 320 panels(modules) wired in strings of 20, so 16 strings.
There could be (this is just hypothetical):
2 combiners / 8 strings each ---> one inverter (two input).
2 combiners / 8 strings each ---> two inverters (one or two inputs i suppose)
4 combiners / 4 strings each ---> four inverters
no combiners / 1 string per inverter = 16 inverters.

I'd call all 320 panels the "array" in all of the above, and the sections going to respective combiners "sub-arrays".
That way the sub-array is an "electrical sub-set", because it is going to separate inverters or inverter inputs or combiners.

It's a 'mechanically integrated assembly' of the components you'd expect.

My opinion: if you have two sets of modules that are sufficiently separated from each other that you need a raceway to support or protect the conductors between them,

Like you said, you have to look at every project, because there might not be any conductors going between strings. (I'm calling a set of modules a string here)

A string is electrically integrated, not mechanically integrated. They are two completely different things and speaking generally there is no telling how they will relate. One could have multiple strings per array or multiple arrays per string. You have to look at the individual project.

Blue for sure, but I really don't get how green would work.
A string could have a + connector in spot X, and the - connector from the opposite end of the string (electrically) could be 1 foot away, or 30 feet, or 100 feet from spot X.
But + or - (or both) have to be fused, so the + and - conductors from every separate string will be both electrically *and* mechanically connected in one or more fused combiner boxes?

Aren't all of the strings connected to a single inverter integrated in a way, in that it turns the whole bunch off if there is a problem with one?

 

[GoldDigger]

There is no simple way for an inverter to "turn off" PV panels in a string unless they have per-panel optimizers or disconnects. All the inverter can do is open it's connection(s) to the string(s).

mobile

 

[PVfarmer]

There is no simple way for an inverter to "turn off" PV panels...

You are of course 100% right.
When I said "it turns the whole bunch off if there is a problem with one? "

I should have said the inverter opens all connections to all strings if there is a problem with any one string.

Here's a pic- I don't get why the four strings are required (if they are?) to have 4 ground conductors to the ground bar of the combiner, when there will be only 1 EGC going from combiner to inverter.
If you had this pictured twice, going to inputs A and B of an inverter (inverter has both + and - connections for A and B)...
when there is only one EGC connection point on the DC side of the inverter- why would you want to connect two EGCs to it when it'll be opening both A and B input connections no matter what?

Pic from here:
http://www.midnitesolar.com/pdfs/MNPV10-12_manual.pdf

 

[GoldDigger]

You need to keep in mind that the wires from the panel frames and racking may be considered as GECs or bond jumpers in some cases rather than EGCs.
That can affect the choice of routing.
In the case pictured, it is also quite possible that the racking of the four strings are not connected or even near each other.

mobile

 

[jjavier]

To be sure:

If our internal AC LV grid is wye grounded, do we need to lay-down a EGC and neutral conductor in each of our AC conduits? Cannot we use a 3P-4W and lay-down only the neutral wire?

regards

Thanks! But, I'm a bit confused by that.
DC too?
How would that be possible here? If inputs A and B on the inverter are separate circuits in 2 conduits, you'd have to ground the combiner twice, which is not right.
(It's a combiner box with only 1 string on each side, so it isn't actually combining, just fusing as pictured.)

View attachment 16742

 

[PVfarmer]

You need to keep in mind that the wires from the panel frames and racking may be considered as GECs or bond jumpers in some cases rather than EGCs.
That can affect the choice of routing.
In the case pictured, it is also quite possible that the racking of the four strings are not connected or even near each other.

mobile

Right. Doesn't it depend on whether the inverter is non-isolated?
I thought with non-isolated inverters, the DC and AC EGC terminals are connected inside the inverter, so it is an EGC coming from the AC side, and that EGC follows the neutral to the SBJ point and bonds there?

And about that pic- so if the 4 strings *are* mechanically connected to each other, you can then tie all 4 together with an EGC and run one EGC from all 4 to the combiner, but if the strings are adjacent to each other but NOT mechanically connected, you'd have to run 4?

To be sure:

If our internal AC LV grid is wye grounded, do we need to lay-down a EGC and neutral conductor in each of our AC conduits? Cannot we use a 3P-4W and lay-down only the neutral wire?

regards

I can't explain it clearly, but I'm pretty sure that's a no, because of the system bonding jumper requirement.

(4) Grounding Electrode. The grounding electrode shall
be as near as practicable to, and preferably in the same area
as, the grounding electrode conductor connection to the
system. The grounding electrode shall be the nearest of one
of the fol1owing:
(I) Metal water pipe grounding electrode as specified in
250.52(A)(1 )
(2) Structural metal grounding electrode as specified in
250.52(A)(2)

(5) Grounding Electrode Conductor, Single Separately
Derived System. A grounding electrode conductor for a
single separately derived system shall be sized in accor-
dance with 250.66 for the derived ungrounded conductors.
It shall be used to connect the grounded conductor of the
derived system to the grounding electrode as specified in
250.30(A)(4). This connection shall be made at the same
point on the separately derived system where the system
bonding jumper is connected.

(6) Grounding Electrode Conductor, Multiple Sepa-
rately Derived Systems. A common grounding electrode
conductor for multiple separately derived systems shall be
permitted. If installed, the common grounding electrode
conductor shall be used to connect the grounded conductor
of the separately derived systems to the grounding elec-
trode as specified in 250.30(A)(4). A grounding electrode
conductor tap shall then be installed from each separately
derived system to the common grounding electrode con-
ductor. Each tap conductor shall connect the grounded con-
ductor of the separately derived system to the common
grounding electrode conductor. This connection shall be
made at the same point on the separately derived system
where the system bonding jumper is connected.

 

[electrofelon]

To be sure:

If our internal AC LV grid is wye grounded, do we need to lay-down a EGC and neutral conductor in each of our AC conduits? Cannot we use a 3P-4W and lay-down only the neutral wire?

regards

Yes you do. All conductors of a circuit must be in the same raceway or cable. 300.3 (B)

 

[jjavier]

Thank you. It´s now clear

Yes you do. All conductors of a circuit must be in the same raceway or cable. 300.3 (B)

 

[PVfarmer]

i was wrong!

i was wrong!

Originally Posted by electrofelon
Would 2 sets of strings each feeding its own inverter but part of the same PV system be two arrays?

I really don't think so....
12 strings, 6 per inverter- that's two inverters and ONE array to me.

Seems like that is in fact two arrays?

Going by page 34 of 168 here:
http://www.pge.com/includes/docs/pd...inspecting_pv_systems_for_code_compliance.pdf

A string is XX modules in series.
An array is XX strings in parallel going to ONE inverter.
The DC combiner takes the XX strings in parallel and reduces the number of source circuits from XX to 1 or 2, depending on # of inverter inputs.

So if your 2 sets of strings equals 4 strings of modules in series, 2 in parallel to one inverter and 2 same to another inverter....that IS two arrays.
But! If those 4 strings go- 2 to input A and 2 to input B of a combiner, and then inputs A/B of ONE inverter- that is one array.

 

[GoldDigger]

Originally Posted by electrofelon
Would 2 sets of strings each feeding its own inverter but part of the same PV system be two arrays?




Seems like that is in fact two arrays?

Going by page 34 of 168 here:
http://www.pge.com/includes/docs/pd...inspecting_pv_systems_for_code_compliance.pdf

A string is XX modules in series.
An array is XX strings in parallel going to ONE inverter.
The DC combiner takes the XX strings in parallel and reduces the number of source circuits from XX to 1 or 2, depending on # of inverter inputs.

So if your 2 sets of strings equals 4 strings of modules in series, 2 in parallel to one inverter and 2 same to another inverter....that IS two arrays.
But! If those 4 strings go- 2 to input A and 2 to input B of a combiner, and then inputs A/B of ONE inverter- that is one array.

So if you use microinverters each panel is a separate array?

 

[jaggedben]

I really don't think so.

Blue for sure, but I really don't get how green would work.
A string could have a + connector in spot X, and the - connector from the opposite end of the string (electrically) could be 1 foot away, or 30 feet, or 100 feet from spot X.
But + or - (or both) have to be fused, so the + and - conductors from every separate string will be both electrically *and* mechanically connected in one or more fused combiner boxes?

Aren't all of the strings connected to a single inverter integrated in a way, in that it turns the whole bunch off if there is a problem with one?

I don't agree with the red. If you have two ground mount racks, with nothing connecting them except some direct burial cable, then if we're using code terms those are two arrays because they're not mechanically integrated. We might call a whole bunch of separate racking structures an 'array' when we're just conversing, but for purpose of code they are separate arrays. For my purposes, if I've got two roof mounted arrays that are connected only by some EMT and some wire inside, those are two arrays, and that's what I call them. They are not mechanically integrated: if the wind rips one array off the roof, the EMT isn't gonna keep it connected to the other one!

To repeat, electrical integration and mechanical integration are not the same thing. Mechanical integration to me means connected with parts that are designed to hold something together and/or in place, structurally, mechanically. I do not consider conductors to provide mechanical integration, whatsoever. They are not designed for that, and frequently require other parts to provide strain relief to hold them in place so that electrical terminations aren't stressed. I wouldn't consider most raceways to provide mechanical integration either, with the possible exception of rigid conduit. Most raceways require parts to hold them mechanically in place, not vice versa.

 

[PVfarmer]

1 If you have two ground mount racks, with nothing connecting them except some direct burial cable, then if we're using code terms those are two arrays because they're not mechanically integrated.

2 We might call a whole bunch of separate racking structures an 'array' when we're just conversing, but for purpose of code they are separate arrays.

3 electrical integration and mechanical integration are not the same thing. Mechanical integration to me means connected with parts that are designed to hold something together and/or in place, structurally, mechanically. I do not consider conductors to provide mechanical integration, whatsoever. They are not designed for that, and frequently require other parts to provide strain relief to hold them in place so that electrical terminations aren't stressed. I wouldn't consider most raceways to provide mechanical integration either, with the possible exception of rigid conduit. Most raceways require parts to hold them mechanically in place, not vice versa.

Panel. A collection of modules mechanically fastened together, wired, and designed to provide a field-installable unit.

For one thing, that written definition makes no sense along with the associated pic.
What it's pointing to in the pic as a "panel" is called a string, and the written part makes it sound like people sell pre-mounted and wired strings.

1- The code terms and pic are confusing. Those "panels" in the pic (which are called strings, by me) don't look and aren't specified as "mechanically integrated".
What if the 2 ground mount racks you mentioned are two strings connected to a combiner with other strings, no cable between strings? The pic show that as a single array.

2- The pic really does make it seem that all of the X number of "panels" going into a combiner are one array. "Panels" meaning (to me) strings of X or XX modules. If I only had the pic to go by, that's what I'd decide.

3- So...are mechanically fastened and mechanically integrated the same thing? Why use both words if they mean the same thing?
Conductors and conduit don't provide *fastening* of course, but what exactly is mechanical integration supposed to mean?

To me, fastened means physically (mechanically) connected, integrated means part of the same system.

 

[jaggedben]

I agree that the code definition of 'panel' is confusing. Nowadays a panel is the same as a module. My guess is that the panel definition dates back to much earlier days when modules were so small that it sometimes made sense to pre-assemble groups of them for installation. That's no longer the case, if it ever really was. I've actually typed up a public input to remove the 'panel' definition.

I also agree that the diagram isn't that helpful and doesn't cover all cases. The code's diagrams, while they have multiplied to cover various cases, still hardly cover them all.

The thing is, the diagrams are really not that relevant. The definitions are more ...er ... definitive. My bread and butter is residential systems where multiple arrays may be on the same string or the same microinverter AC circuit.

 

[PVfarmer]

So if you use microinverters each panel is a separate array?

Um...no?
Going by Fig. 690.1(a)...a string of panels with microinverters is an array.
But! They don't use the terms panel or string! A bunch of MODULES with microinverters connected in what looks like a string is an array with one "inverter ouput circuit"!
:huh:

I've actually typed up a public input to remove the 'panel' definition.

I also agree that the diagram isn't that helpful and doesn't cover all cases. The code's diagrams, while they have multiplied to cover various cases, still hardly cover them all. The thing is, the diagrams are really not that relevant. The definitions are more ...er ... definitive. My bread and butter is residential systems where multiple arrays may be on the same string or the same microinverter AC circuit.

Cool! Thanks for the input/submission.
Since we are both sorta correct, that kinda proves the definitions need focusing.

You could have 20 panels/modules on a garage and 20 of the same on a related house.
That could be 2 or 4 strings, but is 2 arrays for sure. Could be 1 or 2 inverters.

The same 40 panels/modules could be on the ground (no posts, just racking and rails) out back, and be in rows of 10.
Still 2 or 4 strings/ 1 or 2 inverters.
If the 4 rows of 10 are all connected rack to rack *mechanically* by rails, and then I guess meaning all one unit for the wind load calcs, then that's one array.
If the 4 rows are 20 feet apart and NOT connected by rails and all- that's 4 arrays, with the same 1 or 2 inverters.
If 2 rows are connected by rails, then a 20 foot gap, and the other 2 rows are also connected by rails, that's 2 arrays/ 1 or 2 inverters.

I still think it makes sense to ground (by EGC) all the *modules* that are connected electrically to each inverter, to that inverter, seeing as the inverter has only one EGC connection. That's using SMA non-isolated inverters.
So in my "4 array" scenario, the + and - from each string go to the combiner, the EGC goes from string 4 > 3 > 2 > 1 > and one EGC runs from string 1 to combiner/inverter. Does that make sense?

 

[jaggedben]

It's definitely necessary for any array which has modules connected to an inverter to also be connected by EGC to that inverter. There's nothing wrong with having two EGCs connected to one array, or with splicing a single EGC to terminate in two inverters at another location.

 

[PVfarmer]

It's definitely necessary for any array which has modules connected to an inverter to also be connected by EGC to that inverter. There's nothing wrong with having two EGCs connected to one array, or with splicing a single EGC to terminate in two inverters at another location.

"Leaving the vicinity of the array" is still confusing.

When array is defined like this-
Array. A mechanically integrated assembly of modules or
panels with a support structure and foundation, tracker, and
other components, as required, to form a direct-current
power-producing unit.

It sort of implies that every string going into a specific combiner box must be mechanically integrated with every other string landing there to use one EGC (and it sorta sounds like trackers are required...huh?).

I mean...what's the "DC power producing unit" here in the code drawing? It looks like the combiner box (black dot) in the code drawing, as more than one string would have to be called UNITS.

Maybe it should say "panels or strings" for the array definition instead of "modules or panels", since modules and panels are sorta the same thing? :huh:




But!
1 In the top pic here below, if the combiner is "within the vicinity" of the 4 strings, the EGC does NOT have to be run with the (4) conduits for the strings.
2 In the bottom pic, if the combiner is NOT "within the vicinity", the EGC can be run in *one* out of the 4 string condiuts- AS LONG AS the 4 strings are mechanically integrated?
3 BUT, if the 4 strings in the bottom pic are NOT mechanically integrated (say 20 feet apart on the ground, so all within sight but not connected by rails), you then would have to run 4 EGCs to the same combiner, one from each string, (instead of running one EGC string-string and then to combiner as before in #2)?

Say hypothetically "vicinity" means "within 10 feet"- with the 4 strings 20 feet apart the combiner box could be right in the middle of all 4 strings, but only "in the vicinity" of the center two, and not the outside 2!
Again...:blink:

Think of this as an SMA CU-600 or 1000, which has A and B inputs (4 strings each) and A and B outputs (+ and - each), and for EGCs has 4 connection points inside, 2 on side A and 2 on side B.
Also 4 conduiit holes on the botton, 2 on the sides (for EGC(s)) and 2 out on the top. (So it's upside down as pictured)

I'm just unclear on why you would want 4 or even 8 EGCs (from 4 or 8 strings) going into a combiner that doesn't have a ground bar?

 

[jaggedben]

"Leaving the vicinity of the array" is still confusing.

When array is defined like this-
Array. A mechanically integrated assembly of modules or
panels with a support structure and foundation, tracker, and
other components, as required, to form a direct-current
power-producing unit.

In 2017 it will say 'dc or ac', by the way. Which to me indicates that the CMP is not intending to include particular electrical characteristics as part of the definition.

It sort of implies that every string going into a specific combiner box must be mechanically integrated with every other string landing there to use one EGC (and it sorta sounds like trackers are required...huh?).

I don't think it implies that. Remember, many systems don't have combiners or strings, but still have arrays. The definition does not refer to any electrical requirements other than the most general 'power producing unit'. Any mechanically integrated assembly that can produce power could qualify.

A string is an electrical concept, not a mechanical concept.

I mean...what's the "DC power producing unit" here in the code drawing? It looks like the combiner box (black dot) in the code drawing, as more than one string would have to be called UNITS.

Maybe it should say "panels or strings" for the array definition instead of "modules or panels", since modules and panels are sorta the same thing? :huh:


View attachment 16848

That drawing is an electrical diagram. The array definition is not an electrical definition.

But!
1 In the top pic here below, if the combiner is "within the vicinity" of the 4 strings, the EGC does NOT have to be run with the (4) conduits for the strings.
2 In the bottom pic, if the combiner is NOT "within the vicinity", the EGC can be run in *one* out of the 4 string condiuts- AS LONG AS the 4 strings are mechanically integrated?
3 BUT, if the 4 strings in the bottom pic are NOT mechanically integrated (say 20 feet apart on the ground, so all within sight but not connected by rails), you then would have to run 4 EGCs to the same combiner, one from each string, (instead of running one EGC string-string and then to combiner as before in #2)?

No, I strongly disagree with all of that. If you're running circuits in conduits, you need to comply with Article 300, and run at least one EGC with each conduit. (Granting that your conduit could be the EGC, if it qualifies.) Nothing in 690 gives you a clear indication that modifies the relevant requirements.

Again, strings are an electrical concept. (The code term, by the way, is 'photovoltaic source circuit.') I will profess not to understand what you mean when you talk about strings being mechically integrated or not.

Say hypothetically "vicinity" means "within 10 feet"-

I think I may submit a public input to remove 'the vicinity of'...since it's neither precise or necessary. If your circuit is leaving the structure that holds up the modules and going somewhere else, it's leaving the array, and the vicinity. Compare to 690.31 which permits single conductor cable 'within the array'. If you're not 'within' the array, then nothing in the code really permits you not to be following article 300.

Think of this as an SMA CU-600 or 1000, which has A and B inputs (4 strings each) and A and B outputs (+ and - each), and for EGCs has 4 connection points inside, 2 on side A and 2 on side B.
Also 4 conduiit holes on the botton, 2 on the sides (for EGC(s)) and 2 out on the top. (So it's upside down as pictured)

I'm just unclear on why you would want 4 or even 8 EGCs (from 4 or 8 strings) going into a combiner that doesn't have a ground bar?


View attachment 16849

Again, that's an electrical diagram that doesn't show anything mechanical. The critical thing is that the modules and any metal parts supporting them are electrically bonded back to the inverter. If all those modules are installed on a single metal structure whose parts are all electrically bonded, then you could probably be running all four circuits in one conduit with one EGC back to the combiner. If the 8 strings are spread over 25 arrays whose exposed metal parts are not bonded to each other by structural connections, then you'll need 25 EGCs, one for each array, and you'll need to run them with the circuits where they leave the array, and daisy-chain or splice or bond the EGCs together in J-boxes or wherever else they come together so they are all ultimately connected to each other and to the inverter. Any electrician who runs EGCs from box to box and fixture to fixture in other electrical work will recognize the similarities. It is something one learns as part of the craft.

I don't know why you'd talk about a combiner that doesn't have a ground bar. Most ought to have one.

 

[PVfarmer]

Remember, many systems don't have combiners or strings, but still have arrays.

I get no combiner sure, but when a string is defined as a 'photovoltaic source circuit.'...how can there be a PV system without one (or more) source circuit(s)?

No, I strongly disagree with all of that. If you're running circuits in conduits, you need to comply with Article 300, and run at least one EGC with each conduit. (Granting that your conduit could be the EGC, if it qualifies.) Nothing in 690 gives you a clear indication that modifies the relevant requirements.

Then part (F) shouldn't exist, basically?
In what realistic situation would you even want to have a conduit with no EGC in it leaving the vicinity? Leaving the vicinity means going to the next area...WHERE THE EGC MUST BE CONNECTED! :huh:
Also, since raceway/cable is *singular*, it makes it sorta sound as if all of your conductors/EGCs can leave the PV array at only one spot.

(F) All Conductors Together. Equipment grounding conductors for the PV array and structure (where installed)
shall be contained within the same raceway or cable or otherwise run with the PV array circuit conductors when
those circuit conductors leave the vicinity of the PV array.

Again, strings are an electrical concept. (The code term, by the way, is 'photovoltaic source circuit.') I will profess not to understand what you mean when you talk about strings being mechically integrated or not.

http://solardock.com/

I mean having the rails between them like in the big pic top of page. Those look (to me) like strings of possibly 22 panels/modules that are mechanically integrated to each other by rails.
If they were farther apart, with no rails, the strings would be separate arrays. With rails, they are all one array.

The thing is, saying "the EGC must be run with the circuit conductors when leaving the vicinity" about that large of an array...doesn't really work.

I think I may submit a public input to remove 'the vicinity of'...since it's neither precise or necessary.

Please do!

Again, that's an electrical diagram that doesn't show anything mechanical. The critical thing is that the modules and any metal parts supporting them are electrically bonded back to the inverter. If all those modules are installed on a single metal structure whose parts are all electrically bonded, then you could probably be running all four circuits in one conduit with one EGC back to the combiner. If the 8 strings are spread over 25 arrays whose exposed metal parts are not bonded to each other by structural connections, then you'll need 25 EGCs, one for each array, and you'll need to run them with the circuits where they leave the array, and daisy-chain or splice or bond the EGCs together in J-boxes or wherever else they come together so they are all ultimately connected to each other and to the inverter. Any electrician who runs EGCs from box to box and fixture to fixture in other electrical work will recognize the similarities. It is something one learns as part of the craft.

The Connection Unit is equipped with two terminals for connecting the grounding conductors.
Each terminal has two connection points.

Could we say the 8 strings are spread over 24 non-mechanically integrated arrays?
12 panels per string * 8 strings = 96
24 arrays of 4 panels each = 96. (so 3 arrays = 1 string).

I'd call that 8 EGCs. The 8 strings consist of 8(+) conductors, 8(-), and 8 EGCs.
But there isn't an 8 position ground busbar in a CU-1000, so you'd have to splice the EGCs down to 2.

On the other hand, if those 8 strings were all mechanically integrated, going to one inverter, and the combiner was "in the vicinity" of the 8 strings, you would need one and only one EGC going into/leaving the combiner to the inverter, and that EGC would not have to run with the circuit conductors until leaving the combiner.

What I'm trying to get at is that a combiner could be directly in the middle of an 8 string mechanically integrated array, but if it is 20 feet from the northmost and southmost strings, while it is technically inside the array, it isn't really "in the vicinity" of the strings that are 20 feet away, mechanically integrated or not.

I don't know why you'd talk about a combiner that doesn't have a ground bar. Most ought to have one.

They're called "equipment grounding terminals" in the SMA combiners, which aren't the same as a 14 position ground busbar you'd find in a different combiner, that's all I meant.

 

[jaggedben]

Then part (F) shouldn't exist, basically?

My opinion: it doesn't really need to exist but was put there to clarify. Among other things, the code contains elsewhere an exception to having to run the EGC with DC circuit conductors. The CMP, for whatever reason, did not want that to apply to PV. The section could be better worded.

http://solardock.com/

Generally speaking the multiple rows of ballasted systems can be considered mechanically integrated (especially since most of them have ballast between the rows, hold two rows down). That product is a little different, but still appears to qualify with structural members tying the rows together and providing electrical bonding, thus allowing one EGC to be run to the whole assembly.

The thing is, saying "the EGC must be run with the circuit conductors when leaving the vicinity" about that large of an array...doesn't really work.

I don't see the problem with it.

Could we say the 8 strings are spread over 24 non-mechanically integrated arrays?
12 panels per string * 8 strings = 96
24 arrays of 4 panels each = 96. (so 3 arrays = 1 string).

I'd call that 8 EGCs. ...

You keep missing the point and talking about strings when arrays is defined mechanically. Take this system, for example. (Just a random photo I found with google that serves well for my point.) In my opinion there are six arrays in that picture, in codespeak. However there are clearly not the same number of strings per each array, and some of the strings could be split up between arrays with jumper connections connecting them. I can't tell you the best way to wire it without seeing the electrical diagram, but I can tell you that any wiring between those arrays will have to contain an EGC.

I think this isn't really that useful a discussion. As I said, an experienced electrician will have an eye for how to make sure everything is bonded with a proper EGC which is run with the circuit conductors. The code is not intended to be an instruction manual. See 90.1(A).

 

[PVfarmer]

In my opinion there are six arrays in that picture, in codespeak. However there are clearly not the same number of strings per each array, and some of the strings could be split up between arrays with jumper connections connecting them. I can't tell you the best way to wire it without seeing the electrical diagram, but I can tell you that any wiring between those arrays will have to contain an EGC.

I think this isn't really that useful a discussion...See 90.1(A).

Sure there are 6 arrays in that pic. I'd guess the strings are 14 panels each (2 strings/smallest array, 7 strings largest array) but who knows, and yes, if some strings were split between arrays there would be EGCs between those arrays, *but* any wiring leaving both of those arrays going elsewhere will have to contain an EGC also.
So there's the problem, there is no physical way to have EGCs in the same raceway when those EGCs are going in different directions.

What I mean is, if strings were split up between arrays in your pic, you could have some circuit conductors going to a combiner in the barn for some strings in array B, and other circuit conductors going from the same B array to the next array to the south.

And since 90.1 is about safety- a vague rule + assuming people will know what to do doesn't sound terribly "safe".

(Where installed)...makes them sound optional. and "or otherwise"...that is terribly un-specific.

In the interest of being more useful, maybe it should say "same raceway...when those circuit conductors reach the equipment supplied by the array."
That means the same thing, yet is less vague.

(F) All Conductors Together. Equipment grounding conductors for the PV array and structure (where installed)
shall be contained within the same raceway or cable or otherwise run with the PV array circuit conductors when
those circuit conductors leave the vicinity of the PV array.