How Efficient Are Cascading Multilevel Inverters for PV Conversion?

[Bhend449]

I'm curious if anyone can validate the efficiency of a multilevel inverter. I've been thinking about them as a better solution for PV conversion. They seem to have many advantages, but when I look at it, a single phase output can seemingly only draw power from, on average, around half of the panels at a time. (Note, I've seen this circuit both with and without the cap. I suppose it could be implied though.)

I keep looking, and I've found a ton of papers describing these things for grid tie systems. Just as a thought experiment, let's say you have 6 panels that have an output voltage (after MPPT and DC-DC boost phases) of 30 V and each can provide 5 amps. So, I've got a peak voltage of 180V which gives me ~120Vrms. Now, the total available power from the panels is 180V*5A = 900W. But, my output power is 120V*5(?)A = 600W. I know that PVs are current limited devices, so I feel like I'm missing something critical. It seems like most of the time you're not providing all of the available power. Someone please tell me exactly how I'm being an idiot.

 

[Baluncore]

Someone please tell me exactly how I'm being an idiot.

Welcome to PF.
You are not an idiot because you can ask intelligent questions.

You are neglecting to account for the cells employed to subtract voltage and for the duty cycle.
The cap across the cell is needed to store the charge generated while not connected.

The fundamental problem is power factor and series–parallel cell connection.
When the voltage is high, all sources are in series but current is fixed by anyone cell, while maximum current is needed.
When voltage is low you could have several cells in parallel but do not need that much current.

I believe you would be better connecting cells in series, then commutating to produce a square wave at the line frequency. That can then be filtered by a resonant circuit, to remove harmonics and clean up the sinewave.

 

[Bhend449]

Thanks Baluncore! So, I should be able to measure a higher peak current through the load than what's available from a panel, because each one has a cap in parallel. When a switch is turned "on", the panel and cap discharge (or provide current) at the same time and their currents add together. Does that sound right?

 

[Baluncore]

Does that sound right?

Everything sounds right if you provide insufficient information that is not self contradictory.
You have not identified how you will compute or generate the gate signals.

What is the load ?
Is the output going to the grid ? or to a voltage and frequency insensitive load ?
Why does it need to be a stepped approximation of a sinewave ?
A heater or lighting system could use Vdc = Vrms.
A motor or inductive load could use an AC square wave with minimal filtering.

 

[Bhend449]

I'm using an mbed lpc1798 prototyping board, and used MATLAB to figure out the wait times. I'm using bootstrap circuits to drive the gates. The load will be just some dummy load for testing.

Cascading multilevel inverters are supposedly more efficient, have less harmonic distortion, and produce less line stress than filtered square wave inverters. I can provide links to papers if you'd like. (Thats not allowed in some forums though) I'm planning on filtering the output.

 

[Baluncore]

The load will be just some dummy load for testing.

Then why do you do it ?

Each load has it's own requirements, DC or AC, square-wave or sine-wave. Until you specify the application, an optimum design cannot be selected. Papers on the subject are also application specific.

If possible avoid attaching papers as files to posts. Posting links to papers is acceptable on PF, but try to select papers from .edu sites, or papers that are not behind paywalls.

 

[Bhend449]

I'm just doing it for funsies. It seemed like an interesting idea. I like building things, and I thought it would be an interesting challenge to see how high I could get the efficiency.

I'm trying to see how valid they would be for grid tie PV systems, so I'll try something resistive and something inductive. If you have any suggestions I'd be happy hear them.

 

[Baluncore]

For a resistive load you are best wiring the panels in series to give DC = Vrms at optimum current.
If you need AC because of switching specifications then simply use an H-bridge to generate a square wave at the line frequency.

For grid tie systems you need a good power factor. The inverter current into the grid should be proportional to the grid voltage. Your system of stacked inverters appears to do the opposite. Your system could never generate more than one panel's current to the grid.

A more efficient way of delivering energy to the grid is by series connection of panels to produce a high voltage, greater than line Vpk, that then supplies an H-bridge switching converter. The efficiency of a system like that can be better than 97%. That is because the switched inductor behaves like a voltage transformer. It can be designed to dynamically regulate current to the grid, to be proportional to grid voltage and so use all the charge available from the panels. For a 60Hz line frequency it would operate at 60kHz or greater.