Hi Mirza,

I am certainly guessing a bit here, but I doubt if something like LQR is a solution to the problem !

The genesis of the problem (as you have observed for yourself) is the energy being suddenly routed from the flying caps - this is a micro level issue in the context of inverter leg topology and switching. While LQR will treat the entire inverter as a single unit (once again, as observed by you), and hence its ability to rectify the problem at a micro level is to be doubted.

Besides, the LQR is a relatively "slow" control algorithm due to convergence, while what you need is control for the fraction of a power cycle.

I would suggest control that can handle the legs at the level of individual switching - such as some form of sliding mode or hysteresis envelope switching control.

But in any case, I would advise a thorough individual leg level inverter simulation before you converge to any chosen control (LQR or any other alternative). Are you planning implementation of the final setup as well ?

-Sanjay

Dear Sanjay,

I would like to Thankyou for the answer. I found your answer quite resonating with the intuition I had so far. It makes a lot of sense to me. particularly I really like the idea of performing analysis at the level of individual switching instants. I had put a little effort in this direction but looks like more effort is required.

As of now I am convinced that this topology might have better control if we use a non-linear control technique, as you also suggested.

It would be very helpful if you could share some good references for the indicated control structures.

Yes, we would finally implement it on the hardware also.

Hi Mirza,

Off-the-cuff I can't suggest references because I am not working on allied problems at the moment.

All I will say is that there are many references by way of papers and books as well. A good starting point will be the IEEEXplore Digital Library, on which you may search for strings like:

• Flying capacitor topologies
• Current hysteresis control of inverters.
• Sliding mode control of inverters

...and you can expect a flood of references.

-Sanjay

Hi Everyone!

So I think, I figured out the reason of this uneven behavior. Firstly, the integral action inside my controller is not able to catch up with the system dynamics, hence we see a delayed current response, as is apparent from the bump! Hence, it became a driving factor to try an alternative control method which provides better dynamic response.

Secondly, the abrupt switch-over from the DC link to the Flying Capacitor creates a bit of confusion for the controller since this behavior is not modeled and hence the controller does not know how to respond. So it tries to catch up but the response is slower.

Given these issues I decided to implement a controller with MPC. And it proved to be a good gamble, with some limitations.

P.S.: I also tried the LQR approach by using the inbuilt MATLAB funcitons, but the response isn't any better when compared to the Pole Placement controller.

Here is a snap of the MPC reponse.

Thankyou everyone for encouraging me!