Hello all,

I´m what you´d call a "seasoned" automation professional, but I´m not capable of explaining this one. The control gurus around here have helped me before.... perhaps I´m missing something really basic.

The application is a large diesel engine plant on the Canary islands, feeding the whole island. It is the only non-renewable power source for the island. There are several engines, all working in speed droop. We have two engines in particular that are exactly the same, mounting brand new Woodward electronic controllers, both with the same software and with the same configuration. Mechanically, actuator is the same, turbo is the same and both engines just went for on-site maintennance at the same time.

Both engines are set to work on KW-based speed droop. A transducer measures the active power generated by the group and a speed bias is calculated based on the active power. Droop is set on both engines at 4%, rated speed at 500 rpm. So for the 0 to 100% load range we´d have a 20 rpm difference in speed setpoint.

One of the engines works as expected. Once breaker is closed, speed setpoint is reset to 500 rpm and the AGC from the plant starts giving raise pulses to pick up load. As the speed reference goes up, the engine picks up load following the droop curve. For example, at 50% load, we have speed reference 510 rpm, speed bias -10 rpm. At the summing point, the reference is always 500 rpm and we are happy.

On the second engine, where everything is exactly the same hardware/software-wise, the loading curve does something strange. The AGC gives the pulses to raise, but the real power is higher than it should be for any given speed reference. For example, at 50% load, instead of a reference of 510 rpm, we have 509. The speed bias from the power % calculation is working as expected (@50% power, -10 rpm). That 1 rpm difference makes the speed reference to the PID be 499 rpm, which given the rated power, translates to 1MW load extra from where the engine should be.
Here an example of what I mean. The 511..34 is the speed reference manipulated with raise/lower. The -12.5 is the rpm bias from the KW droop curve. Output is the reference for the speed PID:


This also causes asymmetry when a frequency disturbance is injected. If a -2 rpm step is injected, the engine goes up 1MW instead of 2, whereas if a +2 rpm disturbance happens, the engine goes down 3 MW instead of 2.

All in all, we have two "identical" systems that behave differently and I cannot explain why. My best guess is that this 1 rpm offset happens during synchronization/breaker closure, but I am not able to explain it convincingly. Any ideas?

 

Hello

Is that an island mode operation mode ?

I worked in Canary Island in Tenerife sur (Granadilla ) They got 2*frame 6FA GTG

Also if possible can you attach here a single line diagram?

 

@WWSpec,

You didn't say (and it's kind of important), if these two machines have operated "identically" in the past, and now they're not. Or are these new machines being commissioned (or possibly rebuilt engines and/or generators) being commissioned for the first time at this plant? (Because this has happened more than a couple of times to me--I arrive on site to be told the machines are not working properly but WASN'T told they NEVER worked properly since commissioning and they now want them to work properly, also omitting that a few "gurus" had been trying different things over time and some of the schemes were never fully removed when they didn't work, so the machines/configurations really aren't identical at this instant in time, and haven't been for some time. Usually, after studying and scratching my head and my arse for a day or two, a kind soul at the plant will approach me and tell me the real story and then we can get down to understanding what's going on and make some progress.)

Where does this speed reference come from (you mentioned AGC)? What are the factors that go into the calculation of the speed reference?

Is this from a "PMS" (Power Management System) of some kind? If so, who makes the PMS--is it Woodward?

You mention a droop curve--isn't that just a straight line (y=mx+b)? Or are there other factors in that calculation?

Personally, I've never understood the Woodward "reverse" droop scheme; it works, yes, but it's not intuitive in the least. And their GAP ain't--blocks don't tell the reader what's happening inside (a simple block like ADDition is relatively easy to understand without graphical representation), but some of the more involved and important blocks have no "graphical" representation of what's going on inside them at all; one has to read a manual to try to understand what's happening inside the rectangle, and very often inputs and outputs appear on both sides of the block making it difficult to determine what parameters are being passed into the block and which parameters are being passed out of the block.

Anyway, I often find with problems like this (two "identical" machines with "identical" configurations) don't always operate identically. And without being able to understand where the two inputs to the ADDition block are coming from and how they are derived (calculated) it's not possible to say for certain what's happening.

Finally, I have taken brand new Woodward electronic components out of the factory packaging (valve actuators, in particular) and found them to not work properly (shocking, I know--but it's true, and it's also true of other manufacturers as well, it's the nature of today's commoditized business world).

I wish you luck. These kinds of "problems" which aren't fully disclosed and properly explained are very trying.