Gas Turbine Reverse Power Incident

Hello Guys,
Recently we experienced an interesting event at our site. We have a CCPP at our site operating in Island mode. There are two Hitachi H-25 GTs and one ST which make up the combined cycle. There is also a small Tornado GT of 5MW capability. We operate with one H-25 GT on Iso mode and rest of the turbines in Droop mode.
One good morning at 05:20 am, the lube oil pump of our Isoch turbine tripped due to motor trouble. As per defined SP and delay, standby lube oil pump started and took over. Between the 1.5 sec episode of primary pump tripping and secondary pump taking over, the lube oil pressure fell as below as 0.10MPa. The tripping SP is 0.06MPa so there shouldn't have been any problem whatsoever during the auto cut-in of pumps. Well, interestingly, there is a whole lot that happened during these 1.5 sec that has left us with our heads scratching. I'll summarize the SOE below:
  • Droop GT was running at 22.5 MW, Isoch GT was at 20 MW and Tornado was at 4 MW
  • Isoch GT's lube oil pressure was also 0.18MPa, Hydraulic oil pressure was 9.9 MPa and trip oil pressure was 0.45MPa
  • System frequency was at 50 Hz
  • At around 5:21:06, Isoch GT main lube oil pump tripped on overload and standby pump started as per cut-in logic
  • This caused the Lube oil pressure to drop to 0.1Mpascal and Trip Oil pressure to drop to 0.23MPa. Hydraulic Oil Pressure remained above 9MPa
  • At the same time, the SRV lost position from 45% to -1.0375% in one second which resulted in dip in P2 pressure from 2MPa to 1.2MPa. After 1 second, SRV began to recover and returned to its original position in next 0.5 seconds
  • During the SRV closing, Isoch GT started slowing down and shedding all its load to the Electrical grid
  • Isoch GT gave all its 20 MW load and became motorized by further taking 3.7 MW during reverse power phase.
  • Isoch GT remained motorized for approx.. 1.5 sec during which the additional load was taken by droop GT and Tornado
  • Frequency of the network dropped as well to 43.35 Hz which resulted in underfrequency protection actuation and 32 MW of load was removed from the network
  • With 32 MW load removed, Isoch GT normalized and the frequency restored back again to 50 HZ
  • The overall time between lube oil pump tripping and underfrequency actuation was less 2 sec

Before going further, I should mention that we do have reverse power protection through G-60 relay but there is a 5 sec delay before tripping so we never got that far.
It is quite apparent that all this mess happened due to strange response by SRV. Its sudden closing during the first second wreaked havoc. Since hydraulic oil pressure never went below 9MPa, the only cause that could have caused SRV to close is Trip Oil Pressure. But The trip oil pressure went from 0.45MPa to 0.23MPa (The dump valve setting is at 0.14MPa ) so, SRV should never have closed in the first place. The detailed study of trend has shown us that when the lube oil pump tripped, in the first second the trip oil pressure went from 0.45MPa to 0.36MPa . During this, SRV had abruptly closed to zero. Trip Oil pressure further went from 0.36MPa to 0.23MPa in the next half second but during this time, SRV was recovering back to its original position. This is the part which has us in splits. We can imagine that dump valve actuated pre-maturely and caused SRV to close but it should have stayed close if the pressure was decreasing to 0.23MPa. Why did the SRV recover when trip oil pressure was going from 0.36 to 0.23?
I understand there maybe some mechanical issue with the SRV but we had only just started the said turbine 5 days before the event after undertaking its shutdown where SRV was thoroughly checked. Its calibration verified, stroke checked and servo replaced. SRV was working perfectly fine and has been working fine after the event also with no abnormality whatsoever.
Need your guys opinion and expertise on this. Any data you require, I'll be happy to provide.

Can you attach the P&IDs of the Lube Oil and Trip Oil system?

Can you tell which control system is being used on the said turbine?

Can you tell what the load of the ST was doing during this event?

Do you have any fuel flow data for the said turbine?

Do you have any SRV servo-valve current data for the said turbine for this duration of this event?

Do you have the P2 pressure reference data for the said turbine for the duration of this event? (Can you also provide the SRV pressure reference parameters? If the said turbine control system is a GE Mark*, the Control Constant names would likely be FPKGRG and FPKGRO.)

Have you stroked the SRV to some mid-stroke position (having forced 20FG-1 to a logic "1") and then checked to see how fast the SRV closes when 20FG-1 (or 20TV-1) is de-energized? (It should be very quick--VERY quick--to close.)

Do you have data for the GCV of the said turbine during this event?

Can you please clarify the shutdown of the said turbine which occurred 5 days before the event--what work was done on the said turbine? I can't determine if the said turbine was shut down for work on the SRV prior to the event--and if it was, what was the reason for the shutdown? If the said turbine was shut down for work on the SRV, what was the reason for the shutdown--what had the SRV been doing that was suspect? Why was the SRV LVDT calibration checked, the SRV stroke checked (can you also describe what was done during the stroke check and what the exact results were) and why was the servo-valve replaced? OR, was the SRV LVDT calibration checked, the SRV stroke checked and the servo-valve replaced AFTER the event of the said turbine?

I want to stress this point, as I have done so many times before on If "standard" GE-design sequencing is being used for the SRV, the calibration of the SRV LVDT has VERY LITTLE to do with the operation of the SRV. SRV regulator is pressure control--P2 pressure control. The valve is going to go to whatever position is required to maintain the P2 pressure reference. It doesn't matter if the valve has to go to 15.4% stroke or 65.9% stroke or 43.2% stroke to make the actual P2 pressure equal to the P2 pressure reference--it's going to go to whatever position it needs to to make the actual P2 pressure equal to the P2 pressure reference. And, that is going to happen if the SRV LVDT feedback calibration is spot on, or if it's off by 1.3% or 2.7% or 6.9% from the actual valve position--because the SRV regulator is a pressure control loop and the valve is going to be moved to whatever position is required to maintain P2 pressure equal to reference. Full stop. Period.

And--here's the kicker, Makster: Based on the precipitous drop in frequency (remember frequency and speed are directly related) of the said turbine, that means the P2 pressure reference was also dropping precipitously. With the precipitous drop of the P2 pressure of the said turbine during the event, even if the Trip Oil pressure and SRV hydraulic dump valve were conspiring to cause the hydraulic pressure supply to the SRV servo-valve to fluctuate, the flow through the SRV was decreasing because the speed was decreasing because the SRV pressure reference is speed. Pure and simple. And pressure and flow are also both related. As P2 pressure decreases (because the SRV position decreases) the flow through the SRV--and the GCV--was decreasing.

I want to see the Trip Oil P&ID to see exactly where the Trip Oil pressure transmitter is located. Also, can you certify with 100% confidence that there is NO air in the Trip Oil Pressure transmitter sensing line? Ever? Is the design and actual routing of the Trip Oil Pressure transmitter sensing line such that the line never drains back to the L.O. tank when the unit is shut down and no air can ever get into the line? Or are you 100% certain that the line was vented of air prior to the last operation of the said turbine?

Where is the Trip Oil Pressure transmitter in relation to the check valve in the Trip Oil line? (The check valve is usually mounted backwards (or it should be in a standard GE design), and it has a small hole drilled in the flapper of the check valve.) I have found some Trip Oil system check valves improperly installed in the body of the check valve--in other words, they are NOT installed backwards but are instead in the "forward" direction which defeats the purpose of the valve, which is to quickly dump Trip Oil system pressure when 20FG-1 is de-energized to allow the SRV to close very quickly. (When the check valve flapper is not installed properly (backwards), the SRV will not close very quickly when 20FG-1 (or 20TV-1) is de-energized....)

Looking forward to getting the above requested answers and data from the said turbine which was said to be happily provided if requested ("required" was the word which was actually used when saying data would be happily provided...!)
I am sorry I was trying to be detailed but missed some obvious pieces of information. So, let me run them down sequentially:
  • P&IDs are attached
  • Control System is GE Mark-VI
  • ST load was 11MW which increased to 11.5MW only
  • Attached excel file is the trend log for the event. Look for Row # 1133 to find the reverse power entry. You can get fuel flow information and SRV position etc also. The link to the excel file is below:
  • Unfortunately, we did not record any data regarding servo-valve current
  • If you meant the values of FPKGNG and FPKGNO then those are 0.0196MPa and 0 MPa respectively
  • Yes, we have seen the SRV behavior wrt G20FG actuation and it is very, very fast like split second fast
  • You can get the GCV data from the trend file attached
  • To clarify regarding the shutdown we took, it was not only for SRV checking. GT was behaving normally and shutdown was a planned one since its HGPI was due. Performing the stroking and calibration checks of GCV and SRV is a routine practice during all these outages. I must tell you that all the checkings were extremely flawless. Feedback and position were matching exactly. the response time was great as well. This was all PRIOR to the event. Servo replacement was purely management call since we had a new one arrived and the old one was installed for 5-6 years now. There was nothing wrong with the old one and infact I am still keeping it as a spare. During stroke checking, I just take the calibration on manual and give manual commands of 0-100% with 25% increments. The position feedback and LVDT voltages are checked against each command. Nothing rocket science!
  • Regarding trip oil pressure sensing lines, I'm not sure if these have been ever inspected for any air bubbles. I know that no work has been done on these pressure transmitters and their sensing lines for the past two years since we've had no trouble. But still I cannot 100% guarantee what you are asking
  • I assure you that SRV closes very, very quickly when trip oil drains i.e G20FG operates. There is no doubt in that.

Based on this now, I am very interested in hearing your opinion on the subject event. Also, we've never inspected or checked the hydraulic dump valve (GVH5-1 in P&ID). If it is indeed problematic, how can we check it?



Trip Log Live data is pretty coarse; as such it isn't always the best for trying to understand fast events. I have hidden all the irrelevant columns of data in the worksheet I downloaded, and looking at DWATT, DWATT1 & DWATT2 and DFREQ I just don't see frequency decrease you described. I see the opposite--a frequency increase. Which would cause an Isoch unit to reduce load.

Also, looking at the three DWATT signals I see a single instant where the load was negative, and the SRV went negative BUT the unit didn't lose flame.

All very odd bits of very coarse data. The fact that one of the two load transducers reported almost twice as much reverse power as the other at the same instant in time is also unusual, but I have seen similar mismatches in load before and attribute it to differences in sampling/reporting rates of the transducers.

I had also forgotten how much I dislike Hitachi P&IDs. They are drawn large but show so little information they are essentially useless (and I hope you only scanned on page of multi-page drawings).

You must have data from another control system or a "historian" of sorts.?.?.?

Honestly, from the data I see--other than the reverse power and the indication that at some instant in time the SRV went closed (which I find very difficult to believe because that would have most likely put the flame out in one or more combustors resulting in high spreads and flame detector signal dithering for a short time anyway--I think the Isoch unit did what it should have done. It appears that something cause the droop machine to increase its load (but what load is it capable of--I think you said it was at 20 MW at the time of the event) and the Isoch unit was trying to respond--but didn't for some reason. I also see that TNR was changing .... which should not have any effect when the unit is in Isoch control. TNRI should be in control when the unit is in Isoch, and TNRI wasn't changing (at least not per the resolution of the signal in the Trip Log Live file).

Is there some kind of external load/frequency control system at the plant that may have been sending signals to the two H25 units??? I see TNRI is described as "AFC SPEED CONTROL".... Kind of makes one wonder, because that's NOT a typical Mark VI signal longname description. Also, you said the Droop unit increased it's load, which seems to have occurred (because DFREQ increased).

Also, because TNH increased, FPG2 increased--which means if the GCV position was stable the flow through the GCV (and SRV) was increasing.... All strange occurrences.

I haven't really finished reviewing the data, but I wanted to ask the above questions--because I don't see a frequency decrease (at least not around Row 1133; I haven't had a chance to look much further than that). And, I am unable to explain why the Isoch unit didn't drop more load to try to bring the frequency back closer to 50 Hz. Also, the TNRI signal is a real integer--and has not decimal places--not even one. So, it might be changing but we can't see it. (I don't think so--but I'm at a loss to explain why the Isoch unit didn't react quicker to try to bring the frequency back closer to nominal.)

So, I think there's some kind of external load/frequency control AND that very possibly there is some "de-tuning" of the standard GE Mark* Isoch control as well as some non-standard logic for speed and frequency control. It's also VERY odd that TNRI doesn't display any decimal places--at least one, and more like two would be more normal.

Finally, based on the data provided I think you are concentrating too much on the L.O. pump change-over. Sometimes, Trip Log Live data can be very ... misleading, as in the case of the SRV showing it went to -2% stroke. I can understand reverse power--and can even barely justify how DWATT1 and DWATT2 can be so different--but flame was not lost and it doesn't appear there was any spike in exhaust temperature spreads. CPD did drop--at that weird instant in time when the two DWATTs were negative and the SRV was "closed."

But, one more thing--would you please send the alarms from the event, Process and Diagnostic? There should be a Alarm History folder on the HMI that has alarm history for the event in question IF there are no Events shown in the Trip Log Live data file.

And, tell us if there is any external load/frequency control that was active at the time of the event....

Dear CSA,
It was indeed a very peculiar event with many unexplained instances. Looking forward to your further analysis. As far as your queries are concerned:
  • We do not have any external load/frequency control system. An operator basically controls the MW of turbine manually keeping in view the IGVs opening and exhaust temperatures. We do not run our droop turbine on base load and control the MWs by pressing raise/lower speed pushbuttons on HMI ( We do have a load management system which is responsible for shedding load in terms of defined priorities whenever a generator or a combination of generators goes offline)
  • Unfortunately, since no GT tripping occurred, the log captured by Mark-VI is in 1-sec interval. But I assume it would be a very historic coincidence if reverse power and SRV behavior had nothing to do with Lube Oil changeover. Because the said turbine has been working flawlessly since then with no abnormalities
  • The observation that flame did not go off is definitely valid. May be time was too short?
  • I have .bin files of alarms and events which I am attaching. I am not sure if you'll be able to view those:
  • I think I should also explain how we have implemented underfrequency protection. We have six electrolyzers which consume 32MW of load (50% of the entire site). Each electrolyzer has a rectiformer which has its own protection relays. The underfrequency protection is configured in those protection relays. The frequency of 43.35Hz was recorded in those relays which then tripped all six electrolyzers. All this happened so quickly that the frequency increase that you are seeing is probably the response of the system when 32MW was suddenly shed from it
  • My view is that the droop turbines picked up load due to decrease in frequency. This increased the speed error and turbines picked up load
  • Isoch unit did drop the load! In fact it went to negative just to respond to frequency! The unfortunate thing is that a lot happened in that one second of which we do not have the data for
  • Can you explain more on "de-tuning" of Mark-VI logic? If you'll like, I can share the xps file of our Mark-VI logic

You can understand my position here now. I have to fill out an investigative report of this event and have to develop action items to prevent its recurrence in future. And the more I look into the event, more confusing it gets. I would appreciate any sort of leads I could get.
Good evening all!

After i have read this thread and all the elements provided on, and some events i noticed that should get answers like following:

I was looking at Generator field voltage and i noticed that it dropped so suddenly ! ( from 25 VDC to 13 VDC during the " reverse power episode") and then drop again till 4.67VDC!! just after this episode occured " see column 1334 in trend events record file that you attached here.
Also Generator field current dropped from 5A To 3.5 A "during reverse power episode" and
My question is :

-What Is the underexcitation limiter settings of the Generator Control Panel ( it include AVR and Digital relay i guess you got G60 )?

I saw many generators in the field , getting underexcitation and AVR/protection Worked fine and properly to at least send alarms or even trip order to the Generator protection panel and Turbine control system.

I am not saying that can be the cause of getting SRV sequencing , but i was wondering how the generator AVR/Protection can manage such situation.

I can also confirm like CSA said that you should focus also on other aspects of what can be causing/origins and what bad effects can occured in such event .

We do not have actually the single line diagram , we do not know the Power capability for such network.
In island mode there can be sometimes Mvar or Reactive power reverse peaking, and they can be really bad for most of the plant equipments.

Ok you told us that you got sort like PMS ( Power management system or Load shedding Controls ) we do not know without seeing what happen to OTHER GT and ST IF load has been shedded/shared correctly .

Also i notice that as you stated :

GT wich was in ISO mode and "threw on "its load and not rejected it which is different as 52G did not open ...

As you stated that there was also underfrequency occured , about 43.35HZ we do not know ever it is line frequency or generator frequency ??

To conclude:

Without having a single line diagram or getting infos from AVR /PROTECTION Panel, Other GT & ST behaviour during such transient state, i would give these elements that i wrote, trying to support you best i can, for troubleshooting and getting the best solution for a smoother operation of your plant.

Hope this can help,
Controls guy25 aka James