Detection methods for AVR loss of sensing in weak networks


Thread Starter

Yngve S

I am working on my thesis, and one of the item I am covering is the challenge of detecting a loss of sensing voltage to the AVR when the network is weak.

My principal (which is also my employer as I am both a student and a service engineer) is in the business of delivering diesel electric power systems for rigs and vessels with a high need for redundancy (DP class 2 and above).

The main challenge with detecting a loss of sensing voltage is that class requirements does not allow us to solely rely on the AVR itself to report about the failure. It also has a very dangerous impact on the power system as all of the ship's propulsion is powered by frequency converters which *will* trip if the voltage rises above a certain limit. This will render the vessel inoperable which can be very serious.

Some vendors has simply installed a dU\dt-relay which will split the switchboard in two sections if the voltage is rising rapidly. This off course causes a 50% blackout, and is not desirable. Another option which my employer currently is using is based on the effect of the sensing loss, but i cant disclose it in detail here.

There is also the possibility to slow down the voltage rise by allowing healthy generators to absorb more reactive power before powering up the field current, but that is not a solution by itself (although it creates a few hundred milliseconds extra for us to act on).

I have scoured the web for information and best practices about this problem, but I really cant find anything. However, this forum pops up at the top of google search listings in very many other searches that i have, and i have the impression that the knowledge in this forum is quite high.

So do anyone here have any input about this special problem?

Usually in AVR control algorithm, the voltage controller is cascaded with field current controller so the voltage controller output is generating filed current controller set point. in AVR manual mode the operator can directly put the field current controller set point bypassing the effect of voltage controller.

To avoid the effect of losing the voltage reading feedback, you can put limits on the out puts of voltage controller (mini limit &max limit).

Max limit to grantee that the filed current set point will not exceed the limit which lead to increasing the voltage.
Mini limit to grantee that the filed current set point will not be less than the point of stability limit.

This limits can be extracted from generator capability curves or generator V curve.
Hi Hanafy and thank you for your insightful response.

Adjusting the AVR limiters would have been an interesting approach, but there are class requirements that prevents us doing so.

All generators *have* to be able to produce 3x nominal current for at least one second during a short circuit.

That amount of current would require a higher OEL setting than what would limit a loss of sensing voltage rise.
Also, a whole second is very long in respect of a loss of sensing.

We are not able to run i manual mode (due to the weak networks), but there is an option in some AVRs to automatically transfer to field current mode upon a loss of sensing. This would maybe halt the loss of sensing, but it requires internal AVR logic and we are not allowed to rely solely on that (another class requirement).
Have you considered using redundant voltage transformers?

Each of the transformers are protected with their own set fuses with micro switches (or CBs with aux contacts). If inclined, you could take the monitoring one step further by using some relay logic to change transformers if the fuse micro switch (or CB aux contacts) change state.

A bit old school, but simple enough to implement.
Hi Peter,

Redundancy is not a bad idea at all.

It will however require some point of merger, but if that point is close enough to the AVR, the risk of failure will be significantly reduced.

I'm not so sure about extra relays, as they tend to introduce new failure modes.