No Grid Power = No Solar Power

[RegGuheert]

As mentioned, the frequency-shifting approach of shedding microinverters could work for you, Tony. The way that works is the microinverters are programmed such that they drop off-line as different offset frequencies *above* 60 Hz, some closer to 60 Hz and some a bit farther away. That way they are all on around 60 Hz and they gradually drop off as the frequency is raised until they eventually all drop off. That way, the central inverter can control all the microinverters.

Germany has made this feature a requirement for grid-tied inverters in larger systems starting at the beginning of 2012, IIRC. The issue is that the M190s that you have do not have this feature. Since they are designed for the US market, it is not clear that Enphase will have a firmware update which will allow you to upgrade to have this feature. I recommend you ask them and see if that might be available for your array.

 

[planet4ever]

A bit off topic, certainly way off topic from the recent technical discussion, but after reading some of the stuff here it suddenly dawned on me that we have both a grid-tied system and an off-grid system at our home. And with not much special equipment, either. That's because we own an RV that has just sat in our yard unlicensed for the past four years. Everything still works, and I spend a bit of time keeping it that way. The reason it just sits is because I decided I could not, in good conscience, feed it gasoline at the rate of 8 mpg to take it anywhere. The reason it is still here is that I decided it would make a great refuge in the case of emergency. You see, it has two solar panels on the roof, and an extra bank of four deep cycle marine batteries that I added. They can supply most of the rig's electrical requirements, including running the refrigerator/freezer and the hot water heater. (Of course that includes the most important electrical requirement of all, my computer. ) We have a sizable propane tank to run the furnace and kitchen range. The refrigerator can also run off that. Neither the batteries nor the the propane will run the microwave or air conditioner, but those are sort of frills, anyway. Besides, we do have an on-board 4kW gasoline powered generator which will run those units.

So for us, an extended period of no grid power would mean moving out to the RV and living even more on solar.

Ray

 

[RegGuheert]

Nope. Look at the information I've already provided. I've familiar with the way two devices work and they're very similar. There is an MPPT block that keeps the panel loaded at it's max power point, there's a battery charger that provides the three stages (CC, CV, and float) for lead acid battery charging, and there's a DC-DC in between that bridges the gap between the MPPT and charger.

I'm sorry, Andy, but that is very far off of any realistic approach to any MPPT controller. If you try to design and build such as system, you will understand that your approach is not just extremely costly and inefficient, it will result in the need to dissipate kilowatts of power that are best left in the array.

I'll say it again: The Classic Controller is a single DC-DC converter with the PV array at its input and the battery at its output. This DC-DC converter charges the battery under the control of a microprocessor. The voltage ratio between the input and output dictates where on the PV curve the Classic Controller operates. When the battery is in the bulk charging regime, it tracks the PV's maximum power point. When the battery is in CV or float modes, it tracks at voltages above the MPP to find the proper power level to regulate the battery voltage.

 

[AndyH]

I'm almost afraid to jump in with my specific request, but...

My current 6.7kW AC system with 35 SunPower 228w / 35 Enphase M190 needs to be off grid capable. I specifically did not read the links earlier on how to do that "cheap" yet.

My thoughts are that I could add 2kW - 3kW to my current system, and now I'm wondering if it would be better to use a DC system with the Midnight Solar equipment. That would take a smallish battery that will get the whole system operating during the day that I could pull 8kW -10kW at peak solar output (and the batteries would make up the difference, if any) and at night, I would struggle with whatever the batteries can give me.

Your thoughts on my hybrid idea.

I think we'd have to start at the beginning of the design process, Tony, and see how much energy you need to power your 'critical circuits' and for how long.

Cheap is relative and depends entirely on how much backup power you want to have on standby. The refrigerator and CFL or LED lights can be cheap. Adding the hot tub and electric furnace changes the problem in a big way.

Your emergency uninterruptible power supply could be as simple as a battery, and a grid-connected inverter. It wouldn't need additional PV or a separate charge controller. I believe this can be done with an Outback GFX or GVFX inverter, the battery, and whatever manual disconnects code requires. The Outback inverter has the inverter, transfer switch, battery charger, and grid anti-islanding hardware built in. They run at or below ~$1700. The inverter can manage a generator as well. What this system will not do is feed from your existing panels when the grid's down.

Hill linked us to an UPS that uses a Prius battery:
http://www.mynissanleaf.com/viewtopic.php?p=58234#p58234

Homepower Magazine's 'Midnight Special' UPS
http://www.homepower.com/sites/default/files/uploads/webextras/midnight.pdf

Beyond that, you get into full-blown systems. Some example configurations from Midnite Solar:

Grid-tie with battery backup:
http://www.midnitesolar.com/pdfs/gridTie_batteryBackup.pdf

Grid-tie:
http://www.midnitesolar.com/pdfs/gridTie.pdf

Off-grid:
http://www.midnitesolar.com/pdfs/gridTie_batteryBackup.pdf

 

[AndyH]

Nope. Look at the information I've already provided. I've familiar with the way two devices work and they're very similar. There is an MPPT block that keeps the panel loaded at it's max power point, there's a battery charger that provides the three stages (CC, CV, and float) for lead acid battery charging, and there's a DC-DC in between that bridges the gap between the MPPT and charger.

I'm sorry, Andy, but that is very far off of any realistic approach to any MPPT controller. If you try to design and build such as system, you will understand that your approach is not just extremely costly and inefficient, it will result in the need to dissipate kilowatts of power that are best left in the array.

I'll say it again: The Classic Controller is a single DC-DC converter with the PV array at its input and the battery at its output. This DC-DC converter charges the battery under the control of a microprocessor. The voltage ratio between the input and output dictates where on the PV curve the Classic Controller operates. When the battery is in the bulk charging regime, it tracks the PV's maximum power point. When the battery is in CV or float modes, it tracks at voltages above the MPP to find the proper power level to regulate the battery voltage.

I'm ok with much of what you've written here, Reg, except for this:

The voltage ratio between the input and output dictates where on the PV curve the Classic Controller operates.

What I want to see from you is ANY documentation from ANY charge controller that proves that when the power demand from the charger subsystem decreases, that the MPPT subsystem (whether actually doing MPPT or simply PWM the PV input) is purposely moving to different locations on the panel or array's EI curve as you keep stating.

I understand that what you say 'could' be done, but there are other options. What I want to know - with documentation, please - is what IS done. Either you can do that or you cannot - no harm, no foul either way - but a continuing string of "is too, is not" stops now, if you please.

 

[RegGuheert]

I understand that what you say 'could' be done, but there are other options.

There are no other reasonable options. Once you commit to MPPT, other options for handling absorption phase are taken off the table because you already have the best option for regulating in place for absorption and float mode. Any other approach adds cost and complexity and reduces efficiency and reliability.

As mentioned, once you have the converter needed to perform MPPT in place, the other modes really are trivial by comparison. MPPT algorithms are extremely difficult to implement, mainly because it is hard to define the problem fully.

 

[Smidge204]

MPPT algorithms are extremely difficult to implement

...what? Writing a loop that finds the local maximum of a function is pretty trivial.

AndyH linked to a guy's project where he built a MPPT controller based on an Arduino Due. 8-bit microcontroller with 14kb of program memory, 1kb of working RAM, 512 bytes of EEPROM running at 16mhz. The actual algorithm itself is like 80 lines of code, and that includes battery monitoring and charger control. :|

It's a cute little project. Picking through it now in more detail...

Edit: So yeah, the actual hill climbing code is, as expected, trivial. It boils down to:

Do
{
     If (previous_watts >= current_watts)
     {
          delta = -delta;
     }

     pwm += delta;
     previous_watts = current_watts;
}
Loop

Yep... take a tiny step along the function, if it does gown then go the other direction. Incredibly simplistic and, since we know there is only one maximum in the (well behaved) curve, it will find the global max in no time. He uses pulse width modulation to effectively limit the current of the circuit, so "pwm" is the load factor for that (0 to 255). "current_watts" is calculated from sensor input (voltage * current).

We can modify this to a proposed Load Power Point Tracking (LPPT) by specifying a particular value rather than just checking to see if the slope has changed, and changing direction based on if we're above or below the target...

Do
{
     If (current_watts > target_watts)
          pwm += delta;
     else
          pwm -= delta;
}
Loop

Of course a little bit more sophistication might be nice to avoid jittering if the target is between minimum step sizes, and maybe you want to do some kind of quicker scanning with larger steps and gradually decrease step size until you home in on the desired value. Neither of these is difficult to do.

The real trick is figuring out what target_watts should be!
=Smidge=

 

[RegGuheert]

MPPT algorithms are extremely difficult to implement

...what? Writing a loop that finds the local maximum of a function is pretty trivial.

Yes, they are extremely difficult to do well. Like most software, it is trivial to get a prototype working. Basically ALL the vendors out there doing MPPT are either still tweaking their algorithms OR they are way behind in performance.

Here is a test of several MPPT charge controllers done by Midnite Solar: Test of different controllers with identical arrays

So if this function is trivial, then why do some of the products on the market outperform others by 40% some days? A small part of the answer is that the Classic Controller likely has a more efficient power stage. But the lion's share of the answer is that Midnite has spent nearly a decade refining their MPPT algorithm. In fact, their power stage was in place pretty quickly, but the product was about FOUR YEARS late to the market mostly because of the software.

BTW, the same team designed the Outback FM-60 and the Midnite Classic 150, so the Classic was not their first attempt at MPPT.

 

[Smidge204]

So if this function is trivial, then why do some of the products on the market outperform others by 40% some days?

There's more to the sauce than the Tomatoes? :ugeek:
=Smidge=

 

[RegGuheert]

There's more to the sauce than the Tomatoes? :ugeek:

Yeah. Clouds are a huge deal. The edge of some clouds cloud first magnify the light by as much as 25% and then they block most of the light. Then the process reverses when they move off the array.

I think a big part of the Classic's strength is that they can sweep the entire range of operating points in something like 20 msec so it doesn't cost them much in terms of energy production to check the MPPT point.

 

[RegGuheert]

I think a big part of the Classic's strength is that they can sweep the entire range of operating points in something like 20 msec so it doesn't cost them much in terms of energy production to check the MPPT point.

Perhaps my memory fails me. I see 500 msec in the manual. If I can find the original reference, I will post it.

 

[Smidge204]

MPPT algorithms are extremely difficult to implement, mainly because it is hard to define the problem fully.

Yes, they are extremely difficult to do well. Like most software, it is trivial to get a prototype working.

Do do-do do-do~

:lol:

A faster sweep has very little to do with the algorithm. As I said, you could implement some more sophisticated step management to speed things up, or you could take some other approach entirely. You can add all the refinement and complexity you want, but at the end of the day all it's gotta do is find the top of the hill.

But running the code on a faster CPU, or using better current sensors and ADC circuits, would easily influence total performance and none of those have anything to do with the algorithm itself. If anything, the ADC for sensor input is probably the biggest bottleneck... There is no other reason you couldn't find the max power point in under 50ms by my guestimation, and that's with the Arduino.

Do you happen to have their source code handy? I'm willing to bet there's a CS101 level hill climbing algorithm in there.

Edit: And before we get TOO far off track, I just want to point out that you have yet to demonstrate that they use power point tracking to deliberately decrease charge power.
=Smidge=

 

[RegGuheert]

Edit: And before we get TOO far off track, I just want to point out that you have yet to demonstrate that they use power point tracking to deliberately decrease charge power.

No one has proposed any credible alternative.

 

[Smidge204]

Edit: And before we get TOO far off track, I just want to point out that you have yet to demonstrate that they use power point tracking to deliberately decrease charge power.

No one has proposed any credible alternative.

Simple current limiting? Voltage chopper? Whatever older non-MPPT devices use?

Why does it even matter what others have or have not proposed? Argument from Ignorance is not gonna fly.
=Smidge=

 

[RegGuheert]

Simple current limiting? Voltage chopper? Whatever older non-MPPT devices use?

None of those solutions work. They have the following problems:
1) they cannot provide enough current at the beginning of constant voltage mode: Constant Voltage mode begins at the same current level where Bulk mode left off. Since bulk mode was using the DC-DC converter from MPPT and you propose CV mode uses "something else entirely", then you need to provide a solution that provides just as much power as Bulk mode. Nothing other than an MPPT power stage can do this. The result is any other approach will immediately drop out of CV mode and back into bulk mode.
2) The Classic Controller can charge a 12V, 24V or 48V battery from a string of panels. If youn switched to anything else and did not use the same power stage, you would reduce your power into the battery by a factor of 2X, 3X, 4X or more. Makes no sense.

Argument from Ignorance is not gonna fly.

I'm not arguing from ignorance. I have told you, in detail, how they work.

Why do you feel a complex, inefficient solution makes sense when a simple, elegant solution is already in place?