Most Efficient Off-Grid Solar Inverter in the World

Intro: Most Efficient Off-Grid Solar Inverter in the World

Solar power is the future. Panels can last for many decades. Let's say you have an off-grid solar system. You have a refrigerator/freezer, and a bunch of other stuff to run at your beautiful remote cabin. You can't afford to throw away energy! So, it's a shame when your 6000 watts of solar panels end up as, say, 5200 watts at the AC outlet for the next 40 years. What if you could eliminate all transformers, so a 6000 Watt pure sine wave solar inverter would weigh only a few pounds? What if you could eliminate all pulse width modulation, and have absolutely bare minimum switching of the transistors, and still have an extremely small total harmonic distortion?

The hardware is not very complicated for this. You just need a circuit that can independently control 3 separate H-bridges. I have a bill of materials for my circuit, as well as the software and schematic/pcb for my first prototype. Those are freely available if you email me at [e-mail]. I am not able to attach them here since they are not in the required data format. In order to read the .sch and .pcb files, you will need to download Designspark PCB, which is free.

This instructable is mainly going to explain the theory of operation, so you can make this too as long as you can switch those H-bridges in the necessary sequences.

Note: I don't know for sure if this is the most efficient in the world, but it might very well be (99.5% peak is pretty dang good), and it does work.

STEP 1: Theory of Operation

Have you ever noticed that you can generate the integers -13, -12, -11, ..., 11, 12, 13 from

A*1 + B*3 + C*9

where A, B, and C can be -1, 0, or +1? For example, if A = +1, B = -1, C = 1, you get

+1*1 + -1*3 + 1*9 = 1 - 3 + 9 = +7

So, what we need to do is make 3 isolated islands of batteries. In the first island, you have 9 12v batteries. In the next island you have 3 12v batteries. In the final island you have 1 12v battery. In a solar setup, that means also having 3 separate MPPTs. (I will have an instructable on a cheap MPPT for any voltage very soon). That is a tradeoff of this method.

To make +1 on a full bridge, you turn off 1L, turn on 1H, turn off 2H, and turn on 2L.

To make 0 on a full bridge, you turn off 1L, turn on 1H, turn off 2L, and turn on 2H.

To make -1 on a full bridge, you turn off 1H, turn on 1L, turn off 2L, and turn on 2H.

By 1H, I mean the first high side mosfet, 1L is the first low side mosfet, etc...

Now, to make a sine wave, you just switch your H-bridges from -13 up to +13, and back down to -13, up to +13, over and over and over. All you have to do is make sure that the timing of the switching is done so that you go from -13, -12, ..., +12, +13, +12, +11, ..., -11, -12, -13 in 1/60 second (1/50 second in europe!), and you just have to make the changes of states so that it actually conforms to the shape of a sine wave. You are basically building a sine wave out of legos of size 1.

This process can actually be extended so that you can generate the integers -40, -39, ..., +39, +40 from

A*1 + B*3 + C*9 + D*27

where A, B, C, and D can be -1, 0, or +1. In that case, you could use a total of, say, 40 Nissan Leaf lithium batteries and make 240vAC rather than 120vAC. And in that case, the lego sizes are much smaller. You get a total of 81 steps in your sine wave in this case rather than just 27 (-40, ..., +40 vs -13, ..., +13).

This setup is sensitive to power factor. How the power divides up amongst the 3 islands is related to the power factor. That can affect how many watts you should set aside for each of the 3 island solar panels. Also, if your power factor is really bad, it is possible for an island to be, on average, charging more than discharging. So, it's important to make sure your power factor isn't horrible. The ideal situation for this would be 3 islands of infinite capacity.

STEP 2: So, Why Is This So Stinking Efficient?!

The switching frequency is ridiculously slow. For the H-bridge that is switching the 9 batteries in series, you only have 4 state changes in 1/60 second. For the H-brirdge that is switching the 3 batteries in series, you only have 16 state changes in 1/60 second. For the last H-bridge, you have 52 state changes in 1/60 second. Usually, in an inverter, the mosfets are switching at maybe 100KHz or even more.

Next, you only need mosfets that are rated for their respective batteries. So, for the single battery H-bridge, a 40v mosfet would be more than safe. There are 40v MOSFETs out there that have an ON resistance of less than 0.001 Ohms. For the 3 battery H-bridge, you can safely use 60v mosfets. For the 9 battery H-bridge, you can use 150v mosfets. It turns out that the higher voltage bridge switches the least often, which is very serendipitous in terms of losses.

What's more, there are no big filter inductors, no transformers, and the associated core losses, etc...

STEP 3: The Prototype

On my prototype, I used the dsPIC30F4011 microcontroller. It basically just toggles the ports that control the H-bridges at the appropriate time. There is no lag for generating a given voltage. Whatever voltage you want is available in about 100 nanoseconds. You can use 12 1-watt isolated DC/DC's for switching the MOSFETs supplies. The total power rating is around 10kW peak, and maybe 6 or 7kw continuous. The total cost is a few hundred dollars for everything.

It is actually possible to regulate voltage as well. Let's say that running the 3 H-bridges in series from -13 to +13 makes the AC waveform too big. You can just choose to run from -12 to +12 instead, or -11 to +11, or whatever.

One software thing I would change is, as you can see from the oscilloscope picture, the state change timing I picked didn't make the sine wave totally symmetric. I would just adjust the timing near the top of the waveform a little bit. The beauty of this approach is, you can make an AC waveform of any shape you want.

It also may not be a bad idea to have a small inductor on the output of each of the 2 AC lines, and perhaps a small capacitance from one of the AC lines to the other, after the 2 inductors. The inductors would allow the current output to change a little more slowly, giving the hardware overcurrent protection a chance to trigger in the event of a short circuit.

Notice that there are 6 heavy wires in one of the pictures. Those go to the 3 separate battery islands. Then there are 2 heavy wires that are for the 120vAC power.

Comments

[Lee Wilkerson]

Why not just include a schematic?

[Josephs482]

I got a little confused when the algebra came in right at the start. It seems ever so complicated?

[DannyZeee]

So, I definitely agree with your claims of efficiency gains
due to not requiring inductors & capacitors, however, the methodology you
use creates what I see as a possible MAJOR efficiency issue. Using the
switching scheme you employ, at many times the H Bridge 2, and especially the H
bridge 3 batteries connect backwards into the circuit, thereby charging, rather
than discharging them. While this has the desired effect of outputting
the correct voltage for a given step, it also puts the reverse connected batteries
in Charge Mode, which is where I believe the inefficiencies come into play. A
lead acid battery placed under charge, takes approximately 1.4Ah of charge, to
produce 1Ah of output. Therefore, if you are continually charging/discharging
your batteries, you might have substantial losses in the momentary
discharge/recharge cycles?

[MPaulHolmes]

Yes, the charge/discharge efficiency will be an important factor. Some battery types are better than others at charge efficiency. In a car, lithium has around a 5% round trip efficiency loss, whereas lead acid is significantly worse, like you say. It's one of those things where it is just another way of doing something, which will have fewer tradeoffs as charge efficiency improves.

[mmcalli]

Definitely want Lithium cells, perhaps Lithium Titanate--LONG cycle life!

[DannyZeee]

So, another way of doing this, would be to use 13 batteries, but in varying capacities, (Bank 0 would need to be twice the Ah rating of bank 1, which would be twice the Ah rating of bank 2, ..., all the way to bank 13) This would require 13 electrically isolated MPPT chargers, with 13 banks of electrically isolated Panels, and 13 H Bridges, but, would give you your maximum efficiency. (Obviously, some grounding issues would need to be figured out, but, the same grounding issues need to be dealt with in the original configuration as well.) That last part being said, you must leave your battery banks ungrounded in your design? Or are you using some largish TVS diodes that will not pass your battery voltages?

[MPaulHolmes]

I just left the battery banks ungrounded.

[M-Parks]

I have a bit of experience with lead acid batteries, and the 1.4Ah for 1Ah may be true including the overcharge period, but brief charging pulses, especially with the battery below 85% SOC will have very high efficiency, both coulombic and energy. Around 90-95% energy efficiency would be expected. This image originates in a Yuasa datasheet and is discussing coulombic efficiency, but the energy efficiency will follow the same pattern but a bit lower down

[kcraske]

Wow, I'm trying to get my head round the theory however from an electronic point of view what have I missed. As the sine wave is created from successive integers and there is no filtering or shaping why does it not produce a stepped sine wave and if it does are there harmonics flying around they place? Perhaps I should have a look at the schematic
Looks like some keen PCBs.

[MPaulHolmes]

Did you see the oscilloscope picture? I think that picture got buried on the page with all the other pictures. It does step, but the steps are pretty gentle. The total harmonic distortion is actually pretty small. You select the steps so that the area of the steps are the same as the area of an actual sine wave (at least that's what I'm going to do this time. In the oscilloscope picture I didn't do that). It does not cause sensitive equipment any troubles. I've looked at the sine wave that a gasoline generator produces, and the generator's waveform is like 100x uglier. I've powered my laptop with this, refrigerator/freezer, a big milling machine, charged my Nissan Leaf, etc...

[mmcalli]

Pretty solid design you've got going here! Keep looking for these little tweaks and I'll be checking back in when you get boards.

[discostu956]

From a non electronics engineer type who only knows basic electronics, I think this idea seems great. Outside the box type thinking to have separate banks to achieve this.

If you have the separate banks of batteries, does that mean there are separate solar panels to charge the different banks? I'm just wondering if it would possibly cause issues having different panels that might be shaded at times etc without the other panels to even out etc. Or is there a way of branching out the charging from total solar panel output to each bank of batteries?

[aldolo]

it works for sure, but is not practical to have 3 different solar installation for 3 insulated islands. moreover the 3 battery groups discharge unevenly.

[MPaulHolmes]

If the power factor is predictable, you can choose the capacity for each of the 3 islands so the uneven-ness is not an issue.

[RyanMake]

What you gain in discharge efficiency is it lost in charging efficiency?

[MPaulHolmes]

could you expand on that a little? I am a little confused. Do you mean that since some of the islands experience bursts of regen, it loses efficiency from those moments?

[RyanMake]

I was curious that since you have 3 separate power banks that will each need their own charging circuitry dropping to different charging voltages that the charging process in this case might be less efficient than a single bank configuration.

I am also thinking about the blocking voltage of the MOSFETs and limitations of the body diodes. Not a complete thought yet so I will have to think more before I post again.

[MPaulHolmes]

That's a good question. I guess it comes down to the best a person can do with a 12v MPPT, a 36v mppt, and a 108v mppt. A single 156v nominal MPPT may be a little more efficient, but I am not sure. There may be some amazing low voltage designs that are possible.

[Magicpaddler]

Paul


Like the FET
driver solution. A ready make high side FET
driver. How fast does the output of the
DC/DC change? Do you have a scope
picture of the input and output?


My second
question is on switching noise. When a
stage is switching from say negative voltage to zero volts if the switch closes
to provide a zero voltage current path for this stage before the negative
voltage current path switch opens there is a short across the this stage
battery. On the other hand is the Negative voltage current path switch opens
first there is no current path which would cause a voltage spike across this
stage. Do you see voltage transients
across any of the stages? Do you use TVS across stages to reduce transients?


Lawrence

[MPaulHolmes]

Those dc/dcs have about a 3pF coupling capacitance from the input to the output. I've used them in much higher voltage inverter drives (maybe 750vDC bus). I don't have an oscilloscope picture of it, but it is their purpose in life to be high or low side IGBT supplies in switching applications. The dead time is a couple hundred nS to prevent shoot-through. Let's say you have the situation where we have the 9 battery H-bridge. This is the worst case scenario. It is in the state 1H == off, 1L == ON, 2H == ON, 2L == OFF. (so, -1), and let's say you have current going INTO the 1H/1L leg, straight to the 9 battery ground. The 1L leg suddenly opens. Then, the voltage on the lead heading into the 1H/1L leg suddenly goes from 0v to 9*13v relative to the 9 battery ground. The current still has to go somewhere, so it stops going through the 1L mosfet, and passes through the body diode of the high side mosfet for ~50nS before the high side mosfet starts to conduct (since both high side mosfets being on is the 0 state).

I think it does need a very small filter on the output to deal with the dead time noise, which I am going to add on the revision. When I was testing this I had a little inductance on the output just by accident since it was a very long run to my house from the shop. So, I didn't see any spikes on the AC line.