Generator Diode Failures 1

[mgtrp]

We had a diesel generator blow a pair of diodes in the rotating rectifier during synchronising (both failed open circuit). The generator is a 4.16 kV/3 MVA unit with brushless PMG excitation. The immediate cause of the failures seems to be that the generator breaker closed onto its bus with the voltage at only 75% of the bus voltage. We can easily prevent this from happening again with changes to the control and synchronising systems. Looking at the event did raise some other questions for me.

I usually try to keep away from the theory surrounding the interaction between the armature and field in a generator, but I do know that significant voltages and currents can be induced in the field during a short-circuit or out-of-sync close event. I assume that these are somewhat proportional to the current during the event, and the currents during our low-voltage synchronising should have been much less than those during a short circuit and so well within the design ratings of the generator (assuming it was designed for a short circuit event). If this had been a one-off failure, I would probably assume that the diodes might have been a bit tired and think nothing more of it. But, anecdotally at least, this generator has a history of failed diodes in similar circumstances (while synchronising). Is there something else that we should be investigating?

Next, after the failure, the generator voltage gradually decreased across the next 7 seconds to around 30% of nominal, while the AVR boosted the field current until it hit its limiter. I've attached a plot showing this. Does this seem typical behaviour for a generator after diode failure?

Thanks in advance,
mgtrp

 

[catserveng]

Well, an out of phase close like that can do some pretty significant damage to the generator, and possibly the coupling and even the crankshaft. Since you have repeated failures of the diodes, have you done some further checking, like making sure the exciter rotor is good, the surge suppressor is ok, the PMG voltage is not falling off? Have you recorded data similar to what you have from the failure when it doesn't have a problem?

When you say "pair of diodes" do you mean one each of a positive and negative diode? or two of the same polarity? Do the same diodes always fail? Most generators that size have three positive and three negative diodes, is this the case with your unit? What kind of surge suppressor is installed? An MOV type or an older stacked rectifier type?

Depending on many times this is actually happened you may want to give the generator a good set of electrical tests and a good inspection. I had a similar issue a few years ago, the root cause turned out to be a loose jumper connection between the main rotor poles.

Can you provide any specific info on your generator, like who made it, model, how old, how many hours, etc?

Hope that helps, MikeL.

 

[mgtrp]

The close was not out of phase, it was just with the generator voltage too low (about 3 kV, when the bus was at 4.16 kV).

The generator has six diodes, exactly as you described. The pair of diodes that failed were the positive and negative on a single phase of the three-phase exciter output on the rotor. It is not always the same pair that blow, but usually the positive and negative pair from a single phase. I'm still waiting to hear back about the surge suppressor (one thought I had is that it may have failed or been replaced with an incorrect unit, which is why the diodes are failing).

It is an GE-EMD set with a Electric Machinery generator. Unfortunately, this is one of the worst documented generators I've worked with--not even the nameplate is complete! The unit as a whole was installed in the late 70's, but suffered a fire at some stage and was at least partially replaced. Which parts, I'm not sure--some correspondence I have from the 70's appears to reference the same generator serial number, but a different manufacturer.

The failures are more common than most of our diesel generator sets, but not terribly regular. We had two on the same day last week (my description is for the second event, which was the first time we had the DECS-200 AVR properly set up to capture some of the event data), but I don't believe it had happened previously this year. I don't usually work with our diesel generating plants, and they're poorly documented, so I'm relying mainly on anecdotal evidence from others.

For background, the issue with the control system would only occur sometimes--when starting up, the unit control system would initiate a ramp from idle to rated speed, and automatically start synchronising 30 seconds after beginning its ramp. The AVR was set up to turn on just before the unit got to rated speed, with a 20 second soft start. If the unit synchronised quickly, then the voltage would be low when the circuit breaker closes as it's still soft starting. More often, the unit takes some time to come into synchronism, giving the AVR time to reach rated voltage. We are addressing the control system issues.

 

[catserveng]

So is this a multiple unit system and this unit is paralleling to another diesel generator? So you get a breaker close and the the oncoming unit quickly rises to near nominal voltage, so the unit online is basically dragging the oncoming unit up to rated voltage, but based on your data doesn't seem to quite get there. It is a good bet at this point the didoes failed. What is happening to the other unit at this time? I would still recommend a good set of generator tests and inspections since when the breaker closes with that large of a voltage difference there is a good bit of current flowing trying get the paralleled units to the same voltage.

Normally on systems I work with we don't allow synchronization to start until the oncoming unit is above 90% rated voltage and frequency. On a system this old that may not be in place, or was bypassed at some time. Based on your descriptions I'm assuming you have a Woodward SPM-A or something similar? Is it a voltage match type? What kind of sync check relay is in place?

Ideally I would expect that you would want no more than a 5% voltage difference between the units when the breaker closes, and on an older unit depending on a number of factors even that could be too much. I'd say that only blowing two diodes when this happens is lucky, but I can't imagine the surge suppressor surviving that, unless it is open. Have you tested it?

The last EMD units I did of similar size, we also retrofitted to DECS200 AVR's, the units were also retrofitted with 2301E governors driving EGB-13 actuators. These are 900 RPM 60Hz machines, correct? Do you really need a 20 second soft start rate? Most units of this size I set to start at 40% with a start time of 8-10 seconds, unless they are life safety. That usually gives minimum voltage overshoot at startup. Since you have a 20 second soft start time I'm making the assumption this unit isn't required to serve life safety loads as an emergency standby?

Hope that helps,
MikeL.

 

[mgtrp]

This is a grid-connected generator. While there are other units on the same bus, they weren't online at the time. However, your description of the grid dragging the the unit towards rated voltage, and then the diodes failing during this period, is exactly what I think happened.

The generator control system has been modified a lot over the years, and is now used a modern PLC for sequencing, a SPM-A for synchronising, and a 2301D for governing. The SPM-A doesn't voltage match and there is no sync-check relay, which is part of the problem. Voltage matching is meant to be done by the DECS-200.

I think that our next steps will be, as you suggested, to do some general inspection and testing on the generator, paying particularly close attention to the surge suppressor.

We also have a lot of items to address in the control system--voltage matching and soft start rate being two that we had already identified, but there are many other issues that I noted while looking into the diode failures.

One thing that I would like to investigate further is diode failure detection, which is available in the DECS-200, as after the failure the unit continued to run at speed-no-load for about a minute before the operator decided to shut the unit down. I assume that it is normal practice for diesel generators to trip for shorted diodes, is this also the case for open diodes or is it acceptable to alarm only for this condition and have the operator shut down the unit once load has been transferred off the unit?

Many thanks for your comments!

Cheers,
mgtrp

 

[catserveng]

In my own experience with the units I've worked on I have usually had better success using the UEL and OEL settings rather than the EDM function in most cases. Getting the EDM settings to work with both open and shorted diodes requires being able to get enough information from the generator manufacturer to properly input the Pole Ratio. While I have been able thru the years to get fairly good information for most CAT and KATO tail ends, along with a lot of the Baylor (now NOV) models, and have been around enough of them with AVR's that had the EDM function to have settings that seem to work fairly well. I tried a couple of years ago to get info for some EM machines in standby service with no luck. I get fairly regular calls to diagnose EDM trips, that in most cases is just improper settings. It is hard to verify that the EDM is properly working unless you know for sure you have a failed open or shorted diode installed and can compare the "good" and "bad" data.

If you take a look again at the curves you provided you can see what appears to be the excitation levels dropping, likely the result of VAR's coming into the generator to pull up the voltage right after the breaker closes. You may find in that case a properly set Loss of Field or Reverse VAR protection may actually do a better job, at least it has in a number of sites I've worked on. I would expect the unit that size to have a 40 device anyway in parallel applications, especially grid connected. Now here is some comments that may generate some additional responses with a different opinion than mine, we learned a LOT about diode failures a few years ago when some genius decided to reduce the cost of the rotating rectifiers we were using in machines between 500 and 4000 kW. We found that traditional offset mho or impedence type relays seemed to perform well on total excitation failures, but when only one or two diodes failed, many times they did not trip. We did find that the 40Q (reactive power) type relay provided us better protection and reliablity. If I can now days I try to integrate the digital AVR (like the DECS) into the protective relay (we use mostly SEL products these days) and the protection engineer I work with has come up with some good logic that can take the alarm output from the AVR driven by the OEL and UEL functions and take a closer look at the VAR and power levels to better protect the machines in a lot of cases. Depends on how important the end user really thinks his machine is, but it really doesn't seem to cost that much more, but it does require additional effort to get it right.

Hope that helps, good luck in your project. MikeL.

 

[waross]

I tried a couple of years ago to get info for some EM machines in standby service with no luck.
It is hard to verify that the EDM is properly working unless you know for sure you have a failed open or shorted diode installed and can compare the "good" and "bad" data.

Mike: Is it feasible to intentionally open a diode and then short the diode to discover the information that you need to set the EDM properly?
Yours
Bill

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[catserveng]

Bill, I have actually done that on a couple of machines, CAT and KATO supported the effort, on a new machine with no other known issues and some minor adjustments to the protection schemes (40 device settings) used at the test sites, we got some pretty good info to help setup a newer generation of AVR's. However at one site with a generator that had been in service for several years, with a history of poorly defined "problems", the testing had some undesirable/unexpected results, such as excitation going to maximum, other "good" diodes blowing and surge suppressor failures. In one case we damaged an exciter rotor.

So I guess my takeaway from that exercise was as long as you have known good generator and sufficient spares and a situation where the generator causing big VAR swings or tripping will not cause a site problem, then it can be a good way to gather info.

As I apply updated AVR's to systems now days I prefer to use other protections as described above rather than the EDM monitoring for protection.

MikeL