Totem pole pfc stability issues when returning energy to the grid.

[Magnethicc]

Hi everyone,
We are developing a bidirectional totem pole pfc and we are having troubles with stability when returning energy to a "weak grid" aka high impedance grid.

The control is rather simple and is very much like a traditional boost pfc (reference current(t) = Vgrid(t) * Bus voltage error / Vgridrms)
The problem we are facing is when we are returning energy to a weak grid, due to its high impedance the sensed grid voltage rises which then the reference current demands more current (again Iref(t) = Vgrid(t)....)
Which rises the grid voltage more and so the PFC builds up oscillations.

It's like the totem pole pfc will react to every change in grid voltage very fast and so If grid voltage rises we push more current to the grid which grid voltage will rise more and so on and so on...

We tried to filter the sensed grid voltage but this introduces a huge phase shift and PF is highly degrated.

Any one has any experience with such a problem?

 

[KlausST]

Hi,

The problem we are facing is when we are returning energy to a weak grid, due to its high impedance the sensed grid voltage rises which then the reference current demands more current (again Iref(t) = Vgrid(t)....)

I´d say this simply is the wrong algorithm.
Why would one rise the current when the voltage goes higher?

It should be the other way round.

From my understanding there should be a voltage setpoint (nominal grid voltage) ... the closer you come to the setpoint the less current you should push.

Klaus

 

[Magnethicc]

Hi,

I´d say this simply is the wrong algorithm.
Why would one rise the current when the voltage goes higher?

It should be the other way round.

From my understanding there should be a voltage setpoint (nominal grid voltage) ... the closer you come to the setpoint the less current you should push.

Klaus

Since the grid voltage is sinusoidal, when grid voltage starts at 0 and goes sinusoidally towards the peak so is the current so the control rule is voltage goes up, current goes up too to maintain the sinusoidal nature.

Regarding the setpoint, the problem is that the bulk capacitor voltage of the PFC dictate how much current needs to be pushed towards the grid regardless of the nominal grid voltage (load pushes energy to the bulk which then pfc pushes energy to the grid). And what is the nominal grid voltage? Grid voltage levels varies from place to place, load, time of day.

 

[FvM]

An ideal PFC sinks or sources sinusoidal current from/to the grid. Using instantaneous grid voltage as reference waveform creates a negative impedance in recuperation mode and possibly causes oscillations. It has also the disadvantage of copying grid harmonics to the reference current.
Possible solutions:
- using a bandpass
- using a PLL

 

[KlausST]

Since the grid voltage is sinusoidal, when grid voltage starts at 0 and goes sinusoidally towards the peak so is the current so the control rule is voltage goes up, current goes up too to maintain the sinusoidal nature.

This makes sense ... for the V(t) .. the instantaneous voltage.
But you also need to do the RMS voltage approach .. with a much slower time constant .. at least to limit the output RMS voltage.

Regarding the setpoint, the problem is that the bulk capacitor voltage of the PFC dictate how much current needs to be pushed towards the grid

Obviously this does not work out well. There is a limit you have to care for. You can´t push "all the available energy" to the grid - when the voltage is too high.

And what is the nominal grid voltage? Grid voltage levels varies from place to place, load, time of day.

look up the definition for "nominal".

***
Your voltage regulation algorithm has either of two problems (or both):
* too slow RMS voltage feedback
* too low of an output impedance. Your source impedance (source = your inverter) needs to be higher than the worst case grid impedance .. to not get into oscillation.


Klaus

 

[FvM]

My favorite text book about grid tied converter control

Wiley-VCH - Grid Converters for Photovoltaic and Wind Power Systems

www.wiley-vch.de

 

[Magnethicc]

An ideal PFC sinks or sources sinusoidal current from/to the grid. Using instantaneous grid voltage as reference waveform creates a negative impedance in recuperation mode and possibly causes oscillations. It has also the disadvantage of copying grid harmonics to the reference current.
Possible solutions:
- using a bandpass
- using a PLL

Bandbass or low pass tuned to around the line frequency and some more (due to product being sent to places with 50 Hz, 60Hz and due to the fact that we first rectify the input voltage so the band pass will see the sensed voltage which is double the frequency so even 120Hz) will surely produce a phase shift of the signal no?

--- Updated Friday at 8:59 AM ---

This makes sense ... for the V(t) .. the instantaneous voltage.
But you also need to do the RMS voltage approach .. with a much slower time constant .. at least to limit the output RMS voltage.


Obviously this does not work out well. There is a limit you have to care for. You can´t push "all the available energy" to the grid - when the voltage is too high.


look up the definition for "nominal".

***
Your voltage regulation algorithm has either of two problems (or both):
* too slow RMS voltage feedback
* too low of an output impedance. Your source impedance (source = your inverter) needs to be higher than the worst case grid impedance .. to not get into oscillation.


Klaus

How does one increase the inverter output impedance?

 

[FvM]

Bandbass or low pass tuned to around the line frequency and some more (due to product being sent to places with 50 Hz, 60Hz and due to the fact that we first rectify the input voltage so the band pass will see the sensed voltage which is double the frequency so even 120Hz) will surely produce a phase shift of the signal no?

You are asking about "pfc returning energy to the grid". How can it connect to the grid through a rectifier? Band pass or PLL works for the fundamental frequency, not with rectified grid voltage. Please clarify about the topology.

 

[Magnethicc]

You are asking about "pfc returning energy to the grid". How can it connect to the grid through a rectifier? Band pass or PLL works for the fundamental frequency, not with rectified grid voltage. Please clarify about the topology.

totem pole pfc using igbt as the low frequency leg.
Basically an active rectifier.

The measured input voltage of the grid is rectifier using a small signal diodes followed by a diff. Amplifier

 

[FvM]

Still unclear if you are implementing two-quadrant (only energy consuming) or four-quadrant (bidirectional) PFC. Anyhow, if you want a filtered or synthetical current reference waveform, you need access to the 50 Hz grid input. You might be able to reconstruct 50 Hz input from rectified waveform, but why?

 

[KlausST]

How does one increase the inverter output impedance?

It´s your inverter algorithm:
Z = V / I
With: V = V(t) of the grid
and I = your calculated I(t)

--> You should introduce a [factor] when calculating I(t)

Klaus

 

[Magnethicc]

Still unclear if you are implementing two-quadrant (only energy consuming) or four-quadrant (bidirectional) PFC. Anyhow, if you want a filtered or synthetical current reference waveform, you need access to the 50 Hz grid input. You might be able to reconstruct 50 Hz input from rectified waveform, but why?

bidirectional pfc.
We use the input voltage waveform to construct the reference current because we do not have any digital control for the pfc, only analog.
Added complexity is that pfc will work in 50 and 60hz locations

 

[FvM]

I understand that you have at least access to grid input voltage. Post #7 comment about rectified waveform is describing to your present circuit implementation, but no not a principle constraint.

I must confess that I'm designing only digital PFC control since about 20 years, some suggestions like PLL (usually an "all-digital" PLL processing ADC input signal) are unsuitable. Bandpass can be implemented in pure analog design, also with automatic switching or continuous tuning of center frequency. There also analog PLL implementations, but probably involving too much effort.

 

[mtwieg]

Hi everyone,
We are developing a bidirectional totem pole pfc and we are having troubles with stability when returning energy to a "weak grid" aka high impedance grid.

The control is rather simple and is very much like a traditional boost pfc (reference current(t) = Vgrid(t) * Bus voltage error / Vgridrms)

This is meant to make the PFC appear roughly as a slowly varying resistance from the perspective of the grid (a positive resistance, specifically).

When the direction of power reverses, it will make the PFC appear as a slowly varying negative resistance. This can still result in stable operation if the grid's impedance is lower than the magnitude of the PFC source impedance (see the middlebrook criterion). Seeing unstable operation at higher grid impedance is the expected behavior from your control law.

So you shouldn't use the same control law when operating as power source to the grid. I'd suggest adopting the control strategies used by grid tie inverters. I'm not an expert on this, but I'd be willing to bet that from a regulatory perspective this will be necessary as well (even if your system were stable, raising the grid voltage significantly is likely not allowed).

 

[FvM]

My bidirectional PFC designs are basically using constant controller parameters independent of power sign. Limit values specified in IEEE1547 should be implemented, e.g. allowing maximal 110 % of rated grid voltage in continuous operation.

Regarding possible instability, we need to distinguish between control of average quantities, e.g. dc link voltage respectively power fed to the grid and instantaneous quantities like sinusoidal grid current, represented by different control loops.

 

[Magnethicc]

My bidirectional PFC designs are basically using constant controller parameters independent of power sign. Limit values specified in IEEE1547 should be implemented, e.g. allowing maximal 110 % of rated grid voltage in continuous operation.

Regarding possible instability, we need to distinguish between control of average quantities, e.g. dc link voltage respectively power fed to the grid and instantaneous quantities like sinusoidal grid current, represented by different control loops.

In your previous post you said you control the PFC digitally.
Can I ask how do you create the current reference?

 

[FvM]

Can I ask how do you create the current reference?

As sketched above, either with a PLL or an adaptive filter, e.g. a SOGI-FLL (frequency locked-loop based on second order generalized integrator). Single phase PLL scheme below:

In addition to quadrature (cos) NCO signal used in phase detector we'll have an in-phase (sin) signal as current reference.

--- Updated Friday at 10:56 PM ---

SOGI FLL scheme below, both diagrams copied from Teodoresco et al, chapter 4 Grid Synchronization in Single-Phase Power Converters, see post #6

 

[D.A.(Tony)Stewart]

A GTI has a great many technical characteristics to behave as a voltage source like the grid with unity pf current under specifications from utilities, safety agencies and other regulatory bodies. This improves the load regulation errors shared by others on the grid and yet follows the phase and frequency drift of the grid to regulate overall load that increases with frequency from rotational loads. There would be a great amount of disturbances of all types in the short and long term that require sensible detection of what to react to and what to ignore. This is necessary to have accurate references and measurements and transform the error to make incremental corrections, both to protect itself and drive the grid and meet all the specifications.

In order to prevent oscillations and accurate way to measuring grid impedance and dampen the current measured that is injected. There are many types of PLLs and control loops such as inverse Park Transform, Hilbert Transform, and Double Frame of Reference are samples.

A chapter starting p300 and appendix C focus on damping resonance.

https://trinhquocnam.wordpress.com/wp-content/uploads/2013/10/grid-converters-for-photovoltaic-and-wind-power-systems.pdf

 

[Magnethicc]

As sketched above, either with a PLL or an adaptive filter, e.g. a SOGI-FLL (frequency locked-loop based on second order generalized integrator). Single phase PLL scheme below:

View attachment 196096

In addition to quadrature (cos) NCO signal used in phase detector we'll have an in-phase (sin) signal as current reference.

--- Updated Friday at 10:56 PM ---

SOGI FLL scheme below, both diagrams copied from Teodoresco et al, chapter 4 Grid Synchronization in Single-Phase Power Converters, see post #6

View attachment 196097

Thank you very much for the help!
If I may a few more question:
1. Will your digital PFC grid current will be perfectly sinusoidal even with a flat top ac grid waveform?
2. Do you have and can share waveforms of PFC reaction to a grid frequency jump / phase jump?

A GTI has a great many technical characteristics to behave as a voltage source like the grid with unity pf current under specifications from utilities, safety agencies and other regulatory bodies. This improves the load regulation errors shared by others on the grid and yet follows the phase and frequency drift of the grid to regulate overall load that increases with frequency from rotational loads. There would be a great amount of disturbances of all types in the short and long term that require sensible detection of what to react to and what to ignore. This is necessary to have accurate references and measurements and transform the error to make incremental corrections, both to protect itself and drive the grid and meet all the specifications.

In order to prevent oscillations and accurate way to measuring grid impedance and dampen the current measured that is injected. There are many types of PLLs and control loops such as inverse Park Transform, Hilbert Transform, and Double Frame of Reference are samples.

A chapter starting p300 and appendix C focus on damping resonance.

https://trinhquocnam.wordpress.com/wp-content/uploads/2013/10/grid-converters-for-photovoltaic-and-wind-power-systems.pdf

Thank you,
I will make sure to learn from the book as suggested by you and FvM.

 

[FvM]

1. Will your digital PFC grid current will be perfectly sinusoidal even with a flat top ac grid waveform?
2. Do you have and can share waveforms of PFC reaction to a grid frequency jump / phase jump?

1. PLL with NCO (first diagram) yes, 2. (adaptive filter) almost, depending on filter Q.

Due to limited gain of current control loop, actual grid current current has some THD, need harmonic compensation for perfect grid waveform.

2. We did occasionally tests for dip and drop scenarios (EN/IEC 61000-4-11), requirement is to ride through short drops or at least perform an automatic restart.

Requirements for grid frequency variation (EN 61000-4-28) are rather relaxed, max 0.5 % df/f per cycle, PFC can easily track it.

Don't have measurement results available. Many detailed results in Teodorescos book.