With a wide range of load shedding devices ranging for less effective under frequency relays with and without df/dt characteristics designed for quick shedding of loads, there still seems to be a lot of options floating around in terms of PLC based and other intelligent devices. I need to conduct loss of generation based transient analysis studies and evaluate subsequently the various options available to finally recommend selection of the most optimum state of the art load shedding scheme for a industrial plant. Would appreciate advice from friends here who have done extensive work on the subject.
Tek-Tips is the largest IT community on the Internet today!
Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!
-
Congratulations cowski on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!
Load Shedding Scheme 4
- Thread starter nkumarp
- Start date
-
1
- #2
I'm in a large industrial, and we use straight underfreq load shedding. Our generation is only about 20% of total load, so our aim is to keep only the most critical loads online during a loss of utility. Your views and findings on this topic would be appreciated.
vishalmehta
Electrical
One of the load shedding method depending upon loss of generation we have adopted is as follows.
The scada system continuously monitors the plant load and the generation. Delta P(mw), a operator settable value is set which is a value of load which system can take in addition to existing load. It is assumed that machines are partly loaded and there still exists some margin in terms of power for them being loaded to base load. Now whenever, there is a loss of generation, load shedding would take place only if there is a deficit of power inspite of loading the machines to (P + Dp), Where P is the plant load and Dp is the cushion available in the system.
Please let us know whether this is helpful.
- Thread starter
- #4
Under frequency relays, suggested by "gordonl" have found to be less effective, as it has first to wait for frequency to fall to its setting level and then has its own time delay plus intentional delay to allow the relay to ride through a power swings.
The end result is it allows the frequency to fall further below its setting, causing a worsening of the situation, leading then to the need for much larger shedding of load than required. Moreover the relays are pre-wired to trip certain load circuits and does not have any intelligence to know whether these circuits are really loaded to the extent expected. If not the situation gets further worsened, as frequency will continue to fall.
The situation could be improved by using relays with df/dt features. This also has problems as df/dt is a varying value, being subjected to the power mismatch at the instant of loss of generation. This will depend on governor response, the load on the system etc etc. One could work out a df/dt for a worst case, but will not help much. More over the relay needs to filter out harmonics, that causes some delay in its operation
At best the underfrequency relays can be used as a standby scheme to the scheme suggested by "vishalmehta". The scheme suggested by "vishalmehta" also cannot determine which feeders to trip on an intelligent basis. The best that can be done is to pre arrange a trip of certain feeders, without really knowing the extent of load on these feeders. The scheme required should dynamically rearrange the tripping sequence depending on the actual situation of load on each circuit. This is really true for a plant with varying loads, not fixed loads. Each circuit load varies from time to time.
Is there a more intelligent system, even a system that uses neural network engine, that can be trained i.e modeled to read each circuit load, generation loading at each instant and then understand, calculate, think and execute. This way the tripping can be effected quickly, and only a small requisite load and that too the right relevant circuits can be tripped.
Any ideas on any such system that one has used in their plant or have read about it.
Thanks for the responses.
The end result is it allows the frequency to fall further below its setting, causing a worsening of the situation, leading then to the need for much larger shedding of load than required. Moreover the relays are pre-wired to trip certain load circuits and does not have any intelligence to know whether these circuits are really loaded to the extent expected. If not the situation gets further worsened, as frequency will continue to fall.
The situation could be improved by using relays with df/dt features. This also has problems as df/dt is a varying value, being subjected to the power mismatch at the instant of loss of generation. This will depend on governor response, the load on the system etc etc. One could work out a df/dt for a worst case, but will not help much. More over the relay needs to filter out harmonics, that causes some delay in its operation
At best the underfrequency relays can be used as a standby scheme to the scheme suggested by "vishalmehta". The scheme suggested by "vishalmehta" also cannot determine which feeders to trip on an intelligent basis. The best that can be done is to pre arrange a trip of certain feeders, without really knowing the extent of load on these feeders. The scheme required should dynamically rearrange the tripping sequence depending on the actual situation of load on each circuit. This is really true for a plant with varying loads, not fixed loads. Each circuit load varies from time to time.
Is there a more intelligent system, even a system that uses neural network engine, that can be trained i.e modeled to read each circuit load, generation loading at each instant and then understand, calculate, think and execute. This way the tripping can be effected quickly, and only a small requisite load and that too the right relevant circuits can be tripped.
Any ideas on any such system that one has used in their plant or have read about it.
Thanks for the responses.
I would have thought that if one is approaching an underfrequency situation desperate measures are required - mucking about measuring how quickly you are dying ain't gonna save you. If the frequency is going abnormal, and it is going so quickly that standard u/f relays won't save you, I would suggest that you are not going to survive anyway. My view is that you would want to break up into islands and shed load as fast as possible, but not spend inordinate amounts of money on fancy predictive methods (which can only be set by experience, and u/f blackouts are pretty reare experiences). The likelihood of saving yourself from a complete shut down might only be increased by a small amount with expensive (hard to maintain and set) systems over cheaper methods, all for a once in a blue moon event.
I think that dumping more load than necessary is a small price to pay for saving the system from a total collapse, and bugger the lost profits from the lost production!
Bung
I think that dumping more load than necessary is a small price to pay for saving the system from a total collapse, and bugger the lost profits from the lost production!
Bung
Check Beckwith Electric website. They have a generator protection relay that has an option for frequency rate-of-change in addition to standard 81O and 81U functions. Seems I remember a paper by Beckwith on this topic.
There are probably others.
If at all possible, I would use an off-the-shelf stand-alone relay rather than a PLC.
There are probably others.
If at all possible, I would use an off-the-shelf stand-alone relay rather than a PLC.
I working in large power plant connected with three arc furnace. here implimented load shedding scheme 1) with grid low frequency( & time delay) major load disconnected at 33 kv level like ruffing & finishing mill motors & eaf, in this case still we are conntected with power system network (grid) 2) with df/dt (time dealy) + under voltage in this case bus coupler open at 220 kv level ( i.e. one bus connetcted load plant utilites & genrator , second bus connected with transmission lines ,eaf & other major load , both bus interconntected with bus coupler ) .its successfully islanded during the grid failure. my experiance "FCX 103B" relay(ABB MAKE) perfomance excellent during grid failure & abnormal condition at 220 kv level.
For your questation answer is first study the load pattern of plant, second critical load & plant base load (utilites)
requirement third connected genrator load throughoff analysis fourth genator stedy state & transient analysis then you decide which kind of load shedding to be applied.
For your questation answer is first study the load pattern of plant, second critical load & plant base load (utilites)
requirement third connected genrator load throughoff analysis fourth genator stedy state & transient analysis then you decide which kind of load shedding to be applied.
Guest
In fact in case of overloading, the power angle of the Generating station increases. Which in turn tries to reduce the speed and hence the reduction in the frequency. The u/f relays may not protect you in case there is a sudden power dip may be because of some accident in connected client's sites or may be because of envirenmental factors. Under such conditions there is the backup protection scheme which helps in proper power management.
thank
thank
-
2
- #9
Its taken a while to reply, but for what its worth ...
I have developed a scheme that seems to be similar to what you were looking for. It controls generating plant (power output, system frequency, voltage etc.) for the electricity system on the island of Guernsey. We have 125MW of installed plant plus a 60MW interconnector to France. The system also provides emergency load shedding in the event of failure of a generator, or trip of the interconnector. It monitors for this using a signal from the Master Trip Relays. On failure of a generator, the system analyses load conditions and spinning reserve, calculating how much load needs to be tripped (if any). It then selects a suitable combination of feeders from a priority matrix, and trips them. The time from reciept of the Master Trip signal to tripping load is less than 100ms, it needs to be of this order of speed to maintain stability of the remaining generators. The system also contains algorithms for under frequency load shedding that is a little more intelligent than simple underfrequency relays.
The system is built using standard PLC equipment, which we programmed ourselves. I wrote the load shedding parts and some of the generation control parts. Since it was first commissioned in 1999, we have made several extensions. The system now comprises dual redundant main processor nodes, including a data link to our main SCADA system, plus seven remote "data collector" nodes. The nodes are linked using a 1Mb data bus, either using fibre optic or HDSL ethernet.
Hope this is of interest.
I have developed a scheme that seems to be similar to what you were looking for. It controls generating plant (power output, system frequency, voltage etc.) for the electricity system on the island of Guernsey. We have 125MW of installed plant plus a 60MW interconnector to France. The system also provides emergency load shedding in the event of failure of a generator, or trip of the interconnector. It monitors for this using a signal from the Master Trip Relays. On failure of a generator, the system analyses load conditions and spinning reserve, calculating how much load needs to be tripped (if any). It then selects a suitable combination of feeders from a priority matrix, and trips them. The time from reciept of the Master Trip signal to tripping load is less than 100ms, it needs to be of this order of speed to maintain stability of the remaining generators. The system also contains algorithms for under frequency load shedding that is a little more intelligent than simple underfrequency relays.
The system is built using standard PLC equipment, which we programmed ourselves. I wrote the load shedding parts and some of the generation control parts. Since it was first commissioned in 1999, we have made several extensions. The system now comprises dual redundant main processor nodes, including a data link to our main SCADA system, plus seven remote "data collector" nodes. The nodes are linked using a 1Mb data bus, either using fibre optic or HDSL ethernet.
Hope this is of interest.
tomatge, We are working on a system similar to what you describe, but on a smaller scale.
We have a client who will have their own generation on site and will be paralleled with the local utility. They plan to import from the utility a certain amount and generate the rest to meet their campus demand.
Our client needs to be able to shed load when necessary to prevent the power imported from the utility from going above that fixed amount. The utility is quite adamant that this needs to happen within 20 cycles, or else the client can suffer signigicant financial penalties.
We're thinking that maybe we not only need to monitor for discrete generator trip events, but perhaps monitor the import power in real time and react. What if an operator inadvertently lowers the onsite generation, and there is no discrete trip input to react to?
I'm curious about how you (and others) acquire the power information into your PLC and at what update rate. Thank you!
We have a client who will have their own generation on site and will be paralleled with the local utility. They plan to import from the utility a certain amount and generate the rest to meet their campus demand.
Our client needs to be able to shed load when necessary to prevent the power imported from the utility from going above that fixed amount. The utility is quite adamant that this needs to happen within 20 cycles, or else the client can suffer signigicant financial penalties.
We're thinking that maybe we not only need to monitor for discrete generator trip events, but perhaps monitor the import power in real time and react. What if an operator inadvertently lowers the onsite generation, and there is no discrete trip input to react to?
I'm curious about how you (and others) acquire the power information into your PLC and at what update rate. Thank you!
SidiropoulosM
Electrical
Direct load tripping as opposed to underfrequency load shedding may be the preferred solution here. You'll have a logic scheme which will trip pre-specified loads immediately when your generators trip. You can make this as simple or as complicated as you like. The scheme may be armed for certain conditions etc.
Load shedding itself is plant specific, 2 approaches can be made.
1 spinning reserve caluculations of generators.
2 frequency based.
We have after much deliberations are sticking to spinning reserve theory. Complete scada programming is required depending upon the generator droop charecterstics.
Mode of islanding.etc
1 spinning reserve caluculations of generators.
2 frequency based.
We have after much deliberations are sticking to spinning reserve theory. Complete scada programming is required depending upon the generator droop charecterstics.
Mode of islanding.etc
jbartos
Electrical
- Jan 15, 2000
- 6,440
Suggestion: Visit
etc. for more info
etc. for more info
-
1
- #14
![[wink]](/data/assets/smilies/wink.gif "[wink] [wink]")
.jwmccoype
Because load shedding is initiated by trip marks from the proection equipment, analogue power update rates are not critical - in fact in our system a degree of damping of power measurements is actually beneficial. We use standard (but quite high accuracy) power transducers, feeding a 4-20mA signal into DC voltage analogue cards on the PLC.
The Brush system mentioned by walk3641 is dedicated to this sort of application, and as such has a number of features that make it ideally suitable. For example, the equipment has specially designed AC input cards for power, voltage and frequency measurement. It also uses interrupt cards for load shedding to give high speed response. We in fact used to use a Prismic system. The reason that we replaced it with our own is that the Brush system is hardcoded by them, and requires the operator to revert back to Brush if they need to change the specification of the system.
In terms of trip response, our system is quite distributed. The majority of the 100ms response time I quoted is actually data transmission around the network. We could fine tune this to improve speed (e.g. send trip marks and load shed commands on a higher polling rate) but this has not been worth the effort since the response is perfectly adequate for us. On a small system, where I/O and processing may be done on a single node, then response becomes a factor only of processor execution time, and could possibly be brought down to something like 10-15ms.
Because load shedding is initiated by trip marks from the proection equipment, analogue power update rates are not critical - in fact in our system a degree of damping of power measurements is actually beneficial. We use standard (but quite high accuracy) power transducers, feeding a 4-20mA signal into DC voltage analogue cards on the PLC.
The Brush system mentioned by walk3641 is dedicated to this sort of application, and as such has a number of features that make it ideally suitable. For example, the equipment has specially designed AC input cards for power, voltage and frequency measurement. It also uses interrupt cards for load shedding to give high speed response. We in fact used to use a Prismic system. The reason that we replaced it with our own is that the Brush system is hardcoded by them, and requires the operator to revert back to Brush if they need to change the specification of the system.
In terms of trip response, our system is quite distributed. The majority of the 100ms response time I quoted is actually data transmission around the network. We could fine tune this to improve speed (e.g. send trip marks and load shed commands on a higher polling rate) but this has not been worth the effort since the response is perfectly adequate for us. On a small system, where I/O and processing may be done on a single node, then response becomes a factor only of processor execution time, and could possibly be brought down to something like 10-15ms.
Thanks, everyone, for your responses.
One thing that seems to be unique is that our load-shedding objective is mainly to protect the client financially, not necessarily to protect the system physically. (Of course there's lots of relaying, outside my scope, for that). Although, an undefrequency event and the load shedding that might result higher up in the system may be what the utility is ultimately trying to prevent with this financial penalty.
We have used GE Fanuc PLC's throughout this project thus far. We're considering GE's power transducer modules, which connect right to a 90-30 backplane, and claim to re-calculate power nearly every cycle. Anybody out there have any experience with these? Coupled with a speedy CPU, we think we can make the decision in about 20 msec. Another aspect I forgot to mention before is that the client gets their utility power via 5 different feeders, so we'll have measure each and sum those up.
One thing that seems to be unique is that our load-shedding objective is mainly to protect the client financially, not necessarily to protect the system physically. (Of course there's lots of relaying, outside my scope, for that). Although, an undefrequency event and the load shedding that might result higher up in the system may be what the utility is ultimately trying to prevent with this financial penalty.
We have used GE Fanuc PLC's throughout this project thus far. We're considering GE's power transducer modules, which connect right to a 90-30 backplane, and claim to re-calculate power nearly every cycle. Anybody out there have any experience with these? Coupled with a speedy CPU, we think we can make the decision in about 20 msec. Another aspect I forgot to mention before is that the client gets their utility power via 5 different feeders, so we'll have measure each and sum those up.
jbartos
Electrical
- Jan 15, 2000
- 6,440
Suggestion: Visit
for:
SICODIS
Dynamic behavior simulation under constraints, including optimal automatic load shedding.
for: "Rockwell Automation Announces RSPower32TM Version 2.0 Software" that includes the emergency load shedding
etc. for more info
for:
SICODIS
Dynamic behavior simulation under constraints, including optimal automatic load shedding.
for: "Rockwell Automation Announces RSPower32TM Version 2.0 Software" that includes the emergency load shedding
etc. for more info
It seems to me this has nothing (directly) to do with your generators. As suggested above you need to monitor your incomming power and trip breakers depending on how far above the limit of power consumption you are. You could make it fancier with monitoring of feeder breakers to determine how many feeders to trip, but you should be able to do with pre selected breakers with an approximate typical load. 20cycles would give enough time to perform at least two rounds of tripping before time is up. (Time delay required between first trip and next decision) (You may also want to wait a minimum time delay after pickup before first trip for security)
jwmccoype
Our load shedding system is also using GE Fanuc, but 90-70 series. I don't have any experience of power transducer modules, maybe they were not available when we designed our system.
The requirement of your utility seem rather stringent - are they saying that NO power excursion (however small) will be tolerated without financial penalty ? If so, then you'll inevitably have to run generation with some margin of error to allow for short term variations in generation power output that could occur for any number of reasons. The other question is how the utility will measure such short excursions, since this level of time resolution is approaching the realms of high speed data logging rather than utility metering !
Our load shedding system is also using GE Fanuc, but 90-70 series. I don't have any experience of power transducer modules, maybe they were not available when we designed our system.
The requirement of your utility seem rather stringent - are they saying that NO power excursion (however small) will be tolerated without financial penalty ? If so, then you'll inevitably have to run generation with some margin of error to allow for short term variations in generation power output that could occur for any number of reasons. The other question is how the utility will measure such short excursions, since this level of time resolution is approaching the realms of high speed data logging rather than utility metering !
ksv
Electrical
- Mar 7, 2002
- 3
hello everybody,
I am involved in designing an adaptive load shedding system for an industrial plant having 2 Nos. GTG's of 19.2MW each and 1 No. STG of 24 MW and paralled to grid.
The system monitors the inplant generations ,import from the grid and power consumptions at important nodes in the plant.
In case of grid disturbance, df/dt realy in the grid incomer trips.The deficit of power at that point in time is calculated,and plant loads as per priority matrix is tripped.
Tomatge,
Since i am continiously monitoring the generation and consumption of power , the scan rate and accuracy for power measurment matters.Can anybody suggest what could be the criteria??
How do I arrive at the time from reciept of the Master Trip signal to tripping the loads, for my system??
Thanking you in anticipation
Rgds,
KSV
I am involved in designing an adaptive load shedding system for an industrial plant having 2 Nos. GTG's of 19.2MW each and 1 No. STG of 24 MW and paralled to grid.
The system monitors the inplant generations ,import from the grid and power consumptions at important nodes in the plant.
In case of grid disturbance, df/dt realy in the grid incomer trips.The deficit of power at that point in time is calculated,and plant loads as per priority matrix is tripped.
Tomatge,
Since i am continiously monitoring the generation and consumption of power , the scan rate and accuracy for power measurment matters.Can anybody suggest what could be the criteria??
How do I arrive at the time from reciept of the Master Trip signal to tripping the loads, for my system??
Thanking you in anticipation
Rgds,
KSV
- Status
Similar threads
- Locked
- Question
- Locked
- Question
- Locked
- Question
- Locked
- Question
