Prevention of Unintentional Islands in Power Systems with Distributed Resources

>> Houtan Moaveni: Good afternoon, everyone, and thank you for joining the first New York State Interconnection Technical Working Group webinar This is Houtan Moaveni, and I’m the deputy director of New York NY-Sun I’m here today on behalf of my friends and colleagues, Jason Pause and Dave Crudele, the co-chairs of the Interconnection Technical Working Group The topic that we are going to cover today is Prevention of Unintentional Islanding in Power Systems with Distributed Energy Resources, but before we dive into the technical content, a few logistics items that I would like to discuss Everyone is in listen only mode, and you can ask a question by using the QA panel on the right hand side of your screen We will have about 45 minutes for our speaker to present, and this leaves us an approximate like 30 minutes for QA at the end of the presentation The webinar is being recorded, and it will be posted on the Interconnection Technical Working Group webpage So with that, I would like to go ahead and introduce our speaker Our speaker is Dr. Ben Kroposki Ben is the director of Power Systems Engineering Center at the National Renewable Energy Lab, NREL, where he leads NREL’s research and design, planning and operation of electrical power system Ben received his bachelor and master in electrical engineering from Virginia Tech, and a PhD from the Colorado School of Mines His expertise is in design, testing, and integration of renewable and distributed power systems, and has more than 115 publications in these areas As an IEEE Fellow, Ben was recognized for his leadership in renewable and distributed energy resources integration Today, Ben will discuss basics around unintentional islands and techniques for detecting and preventing islands He will discuss [inaudible] connection standards and the testing procedures as well as new results conducted by NREL and other national labs Finally, he will talk about the future if anti-islanding protection So with that, I would like to turn it over to Ben to kick off the presentation Ben? >> Benjamin Kroposki: Thank you, Houtan, for the introduction, and good morning, everyone So today, I’m going to give a presentation on prevention of unintentional islands in power systems with distributed resources, and one little additional background on myself I was involved with the original development of IEEE 1547 as a member of the working group I was also the secretary for IEEE 1547.1, which is the conformance test procedures, and I was also the chairman of IEEE 1547.4, which is the guide for operation and design of intentional islands with distributed resources or micro-grids So today, we’ll give a presentation for about 45 minutes, and then we’ll take questions at the end Okay, so our presentation outline First, we’ll talk about various types of islands in power systems Then we’ll move on to issues with unintentional islands, discuss a bit about the methods of protection against unintentional islands, and then move on to some of the testing procedures that are used for evaluating equipment for unintentional islanding Then we’ll talk a little bit about the probability of unintentional islanding, and then move on to the future of anti-islanding protection At the end of this presentation, there is a number of pages of references, so additional material if you’re interested in specific topics that you can follow up on So before we begin into the presentation, I just want make sure that people are straight on terms, and for the most part, these are out of the IEEE 1547 standards, but it is helpful to make sure that you recognize when we talk about these, what we’re really talking about

So things like area electric power system, that’s really in layman’s terms the utility system Local electric power system, that’s a lot of times talking about the electric power system on the other side of the point of common coupling And then point of common coupling, this is where distributed generation would connect to the power system In IEEE 1547, they decided on the term DR, meaning distributed resource, and in this case, distributed resource could mean distributed generation or distributed energy storage or distributed energy resource In newer versions of IEEE 1547, so the one that’s being developed right now, they are getting ready to make a change from DR to DER, and the reason for that is mostly because DR also means demand response, and it gets a little confusing So in the context of this presentation, DR will mean distributed resource or distributed energy resource And then the last item on here, anti-islanding, which is non-islanding protection It’s the use of relays or control to prevent continued existence of an unintentional island These words, unintentional islanding, non-islanding, and anti-islanding, kind of get used quite a bit across these slides So as we take a look at island definition, so what is an island? It’s a condition in which a portion of the area electric power system is energized solely by one or more local EPSs through an associated point of common coupling So if you look at the figure on the right hand side, you can see a breaker would potentially open, and since that has distributed resources and load in there, you could form an island Now inside of 1547 they mention both intentional islands or planned islands and unintentional islands, so we’ll talk about those As we look at intentional islands, these are also known as microgrids, there’s a variety of different types of intentional islands, but the key thing to recognize with intentional islands is they are intentionally planned, so they have distributed resources and load They have the ability to disconnect and operate in parallel with the area electric power system They could include portions of a local electric power system or portions of an area EPS But again, the key feature is that they’re intentionally planned You see in the figure, you know, a local facility microgrid, maybe a building, whereas a larger utility microgrid may include parts of distribution circuits There’s also two standards that are in development right now, P 2030.7 and P. 2030.8, which are covering microgrid control and design IEEE 1547.4 is the standard I mentioned earlier, design operation and integration of intentional island systems But we’re going to spend the rest of the presentation here today talking about unintentional islands And so what are the issues with unintentional or non-planned islands? First off, personnel safety Unintentional islands, if they’re operating, could cause hazards for utility workers, if they’re assuming that power lines are not energized, for example, during restoration So if you have distributed generation that are accidentally maintaining the power and voltage live online, that’s a potential safety hazard There’s also the potential for overvoltage, both transient overvoltages due to the rapid loss of load If a breaker opens and you have generation on the islanded side, you may have a very quick overvoltage condition occur because of the loss of load on that In addition to that, there’s also ground source issues Ground faults can present and result in voltages that exceed 173 percent on unfaulted phases So those are things that need to be addressed in terms of the grounding for distributed generation devices, when you look at how – at potentially how you could create islands in those systems Another issue is reconnection out of phase So one thing that could potentially cause an island is the opening of a breaker or a recloser If you reconnect that system back to the grid, potentially the grid and the island are out of phase with each other, and that could cause a transient torque applied to the motors, which could result in damage or failure So that is a potential issue And then power quality within the – any unplanned island, your distributed generation

may not be able to provide suitable power quality for the loads And then finally, protection issues If you have an unintentional island, you may not be able to be providing enough fault current, for example, to operate fuses or overcurrent relay protection inside the island, and so your protection systems may need to be adjusted There’s a variety of references there you see on the bottom, and then listed at the back end of the presentation Just to cover some highlights with understanding your DER or distributed resources, there’s basically three different kinds You may have a synchronous generator, and typically, these are voltage source devices, something like a diesel generator, maybe a natural gas powered engine-gen set, realizing that those can definitely support island grid operations Induction generators, typically, they’re not able to support an island, and this is because of the need for reactive power to supply on an induction generator Of course, if there’s sufficient capacitance, like capacitor banks inside that potential island, that needs to be evaluated And then finally, inverter-based distributed resources This is something that’s becoming much more prominent with things like photovoltaic systems or energy storage systems, where you have an inverter-based system Now typically, in general, you need a grid-forming inverter to provide that support, so a voltage source-based inverter system A lot of times now these inverters are operating in a current control mode, and so they aren’t necessarily able to provide a voltage source, but are phase lock looped into the AC power system So moving on to specific requirements around unintentional islanding, so IEEE 1547 section 4.4.1 gives the unintentional islanding requirement, and it’s stated here, for an unintentional island in which the DR energizes a portion of the area EPS through the PCC, the DR interconnection system shall detect the island and case to energize the area EPS within two seconds of the formation of the island So the key point here is realizing that it needs to detect and cease to energize within two seconds for the requirement I wanted to throw up a little bit of background material on where did this requirement come from So prior to 1547, there was an earlier interconnection standard focused on PV systems called IEEE 929 This has since been withdrawn and replaced by 1547, but if you go back and take a look at some of the requirements in that document, you can see, for example, that if they gave a couple of conditions, that the inverter must recognize and then disconnect So one of the things was a ten cycle or less disconnection, and that was – it’s a really fast disconnection time, but it’s also, as you can see from the requirements, really [inaudible] power mismatch is pretty great And then if the power mismatch is within a certain nominal level, it’s within two seconds, so that two seconds is kind of carried over from early PV interconnection standards If we look at 1547 and go through some of the various drafts, typically, the drafts before draft 5 had this two second requirement in there There was actually some points in time during draft 6 and 7 when we bumped that up to 10 seconds, and that was actually recommended a lot by synchronous generator manufacturers that weren’t using active anti-islanding techniques, because they recommended that value And then drafts 8 and beyond, the requirement came back to two seconds Two seconds a lot of times is decided as a fast enough that people think that they feel comfortable with that But again, it’s not fast enough to get inside instantaneous recloser settings, so if a recloser opens and then recloses, it’s going to do that within the two second timeframe Inverters, on one hand, especially if they’re used in an active anti-islanding technique, definitely can usually meet the two seconds, and oftentimes much quicker But we’ll get into some of the discussion on this later on Okay, so addition – in addition to the requirement, there’s a footnote in 1547, and the footnote is actually very important to read, because it gives some suggestions on how to meet the requirement So for example, you’re able to meet the requirement if your DR capacity is less than one-third of the minimum load of the electric power system, so the DR is small in relationship to the local load If the DR contains minimum or reverse power flow protection, if the DR is certified to

pass an applicable non-islanding test, and if the DR contains other non-islanding needs, such as forced voltage or frequency shifting, transfer trip, or governor excitation control, and we’ll talk a little bit more about these as we go on in a couple of slides here So first off, that requirement that you would be able to meet the requirement if the DR is less than one-third of the minimum load So basically, what this means is that the DR size is relatively small with respect to the local generation, and there’s a couple of references and work that’s been done on looking, that if you had this kind of ratio, sort of three to one, which is pretty high, that you would be able to not maintain a stable island operation for both induction and synchronous generators It basically showed that the pre-island condition, when it’s approach three times the generation, you can’t really support the load And this is a key thing that is often difficult to determine, is what the actual loads around distributed generation are, but basically, this is saying if you have a small DR in relation to the load, that you won’t be able to support an island condition So as we move on to additional methods for protecting against unintentional islanding, there’s a variety of different types listed here We’ll go through these in more detail Reverse/minimum import power relays, passive anti-islanding detection methods, active anti-islanding detection methods, communication-based anti-islanding detection methods, and some ones that are under development, and I’ll talk a little bit about each of these But if you take a look at the references there, 10 through 37 in the back, there have been – there’s a lot of information available on a variety of different anti-islanding techniques, and I would recommend browsing through these references if you’re interested in more details on any one of these So first off, taking a look at things like minimum or reverse import/export relays What that means is you basically set a relay, if you look in the figure where that red box is the breaker, you may have some relay functionality attached to it These numbers, 81, 59, 27, and 32, are relay functions, 32 is the reverse power, and in this case, if the power is ever reversed onto the utility grid, that would trip that particular breaker This is for what they – utilities typically call a non-export type condition And if you’re set to never inject power back onto the grid, this is typically used for that case Now a lot of times this is when you have a large local load If we move on to passive anti-islanding techniques, basically, this – passive means that you’re building your anti-islanding protection around over and under-voltage and frequency trip point settings So if you take a look at the charts there – sorry, they’re small, but those are at a 15 – the new 1547, with the clearing times based on various voltages and frequencies, and then the new clearing times that will be adjustable So in terms of functionality from a relay protection setting, there are things like the 81 over and under voltage, 27 and 59 are over and under frequency trip points, and if your DR is outside that particular window, then it’s basically going to trip Now when you use a passive anti-islanding technique, there is basically what they call a non-detect zone if your voltage and frequency is being maintained within those areas And so if you are able to create a stable island condition, but your voltage and frequency don’t go outside the normal operational areas, passive anti-islanding can be fooled That being said, it’s very difficult to maintain that matched real and reactive power setting with your distributed generation in there The big thing here is that the new amendment to 1547 allows for adjustable clearing times, and you can see some of these clearing times are opening up much wider For example, in the table, if you looked at the voltage between 45 and 60 percent, the clearing time used to be 1 second Now you can adjust that all the way to 11 seconds And this is accommodating things like low voltage ride through capability, but that’s also going to open up your passive anti-islanding windows Additional passive anti-islanding techniques include things like rate of change of frequency, looking at voltage or current harmonics, looking at things like a phase jump when all of a sudden an island is formed because you do typically have a load mismatch You’ll see that in the voltage or current outputs

So moving on to active anti-islanding techniques, so there’s a list – quite a list here of different anti-islanding techniques These are all active And when we say active, what we mean is that the DR is basically attempting to move the voltage or frequency around on the grid, and then detect whether or not the grid has changed So typically, these work best if you have a very stiff grid That means the DR in relationship to the size of the grid is relatively small, and the DR will not be able to move voltage or frequency around But active anti-islanding techniques in general are trying to create a disturbance When it moves into an islanded mode, it’s actually able to move voltage and frequency and then typically will push faster until it pushes outside those trip windows So as we move onto communication-based methods of anti-islanding, there’s a variety that we’ll talk about quickly First, power line carrier, and typically, this is employed where you use a permissive run signal at all times on the grid, and then when the signal goes away, the DR recognizes that and ceases to energize the circuit There’s things like impedance insertion where you may have a capacitator bank located near your DR that’s remotely controlled, and if you go into an island condition, and you can add those capacitors and cause the over/under voltage protection to activate And then one that’s also used quite a bit, especially for larger scale systems, is what’s called direct transfer trip, and we’ll talk about that on the next slide So direct transfer trip provides a communication signal from the area electric power system, such as to a component, maybe like a feeder breaker or a sectionalizing device, back to the DR or some other device on there The key here is that you have a actual dedicated communication channel This is actually very well-known and used The biggest challenge here with using direct transfer trip is the fact that you do need some type of dedicated communication, and that means that you’re typically installing some type – if you take a look at what I had listed here from the PG&E interconnection requirements, you could use things like a direct fiber link, licensed microwave link, or some other type of leased line The biggest challenge is that these dedicated fiber or other communication methods are often costly to install and operate So for a large scale system say on the order of 20 [inaudible] of distributed generation, you may be able to absorb the cost of a communication system into that project For smaller systems, it’s going to be difficult to be cost competitive to allow that kind of a system to be installed, if they don’t already have some kind of communication mechanism So I’m sure we’ll talk about this a little later on in the questions An additional method that’s under development, this is one that looks at the phasor measurements, both at the distributed generation and at the bulk power grid, and basically uses phasors to understand if the islanded condition occurs So if they are interconnected, you’ll have no phasor difference between the voltages on the system As soon as you have an island, you’ll probably have some drift there because of different load and DG requirements, and you’ll start to see some kind of voltage differential there And that would allow you to be able to detect this Again, for this type of system, you will need some type of communication, in addition to the phasor measurements Okay, so that kind of talks about the variety of different techniques used to detect and mitigate unintentional islands I’ll talk now about – a little bit more about standardized testing that’s done in this area So inside of 1547.1 there’s a unintentional island test that basically defines how pieces of equipment are tested to evaluate for unintentional islanding capability The thing to realize about this test, it is only done on a single inverter or a single distributed generator in this case, so the – in the diagram, you see EUT That stands for equipment under test And then that’s connected to a load bank In this case, it’s a RLC load that basically matches to real and reactive power And to conduct the test, you match the RLC load bank with the output of the device under test until you have basically no current flowing through that switch S3 When S3 – at that particular point, you can open that switch and see how long you

can maintain that island between the load bank and the equipment under test A couple of things just to realize on this is that you do have basically what is considered the worst condition that you can create to understand how well the anti-islanding techniques work So this RLC load is very exactly matched to the output of the unit under test There’s a couple of things to realize about that It’s done at 100 percent, 66 percent, and 33 percent of rated power It’s also quite difficult to detect exactly when you have that match condition, so it’s done at increments – in one percent increments around 95 to 100 percent of the loading conditions The other thing that’s mentioned on this slide is something called a quality factor The quality factor in the test is now equal to one, and that basically is equivalent to a worst case power factor of .7 So when we were designing and developing these tests, we wanted to create basically a resonant circuit that would try to hold the device at a 60 Hz, or if you were doing a 50 Hz for a European standard, system So basically, what this does is it sets up what’s called a resonant tank circuit That RLC load is very tightly matched It also has got this ability, using the LNC, to create a way to try to hold that frequency steady, so that the devices can’t necessarily push it around In older tests, like in 929, we were using a quality factor of 2.5 Based on a lot of testing we did in the early nineties, we basically came to the conclusion that we were seeing very similar results using a quality factor of one, and that was much easier to actually set up and run that particular test So that’s why the standard now has a Q of 1 As we look at a couple other tests inside of 1547.1, there’s one for synchronous generators Again, the key here is to match real and reactive power, and then get zero current flow across switch S1, open that up And for synchronous generators, this is done at things like minimum load at unity power factor, maximum load at unity power factor, real load at rated power factor leading and lagging So a variety of conditions And then also there’s a test in there for understanding reverse power flow So this is, for example, if you have reverse power flow protection in the case of relays, you’re able to set up and test and evaluate those Again, these are the standardized tests in 1547, developed over a number of years, and have been used successfully for quite a few years And those are also integrated into UL 1741 for actual certification processes Again, the big thing is to realize is that that’s on a singular distributed generation, and some of the challenges we have when running those tests is the ability to match real and reactive power extremely precisely Even in a testing laboratory, it takes a lot of work to get that very precise, and cause these island conditions And the other thing, like I mentioned before, is that it is only done on one test – on one inverter So what I’m going to cover next really covers advanced testing to resolve some of the issues that I just brought up So first, I’ll just mention at NREL we do a lot of this work in our energy systems integration facility We have the capability to test and evaluate equipment up into the megawatt range for these types of tests And we’re also able to integrate that with a lot of additional capabilities And we’ve looked at islanding tests for things like PV systems, wind systems, energy storage systems, electric vehicle type systems, over a large range of distributed generation devices And some of the things that we’ve worked on and tried to work a little bit on include improving those tests So one of the things that I’ll just mention, one of the very big challenges is when you have a large scale singular device, like a megawatt or greater than megawatt, you need an RLC load bank that has all of those components rated at that power level And one of the things that we did was some advanced testing looking at could we replicate the load bank and grid conditions in what we call hardware in the loop or hardware in the loop And what this allows us to do is not use a physical load bank, but actually simulate the load bank conditions So there’s some references on this particular type of testing, but we were able to basically

replicated and validate the test that we were – that we typically do in full hardware with the ability to replicate that in software And it also actually allowed us to much more finely tune these resonant tank circuits, so we were able to show that using these simulation approaches, that we were able to replicate the anti-islanding tests and results One of the drawbacks on this is it may not work on all active anti-islanding methods, because a variety of different anti-islanding techniques may actually use the load as a reference point If we go on to additional things, one of the questions that often comes up is what about multiple inverters? Obviously, it’s rare – it’s getting rarer now that you see singular DG installations So we’re starting to see much more multiples of inverters So both Sandia and NREL have done a variety of tests in this space So Sandia, you see the reference there, examined four inverters at a single point of common coupling, and demonstrated that multiple inverters still met the two second requirement that’s in the standard Some recent testing that was just released here in July of this year, so last month, was testing that NREL completed with SolarCity and HECO to examine two additional things What’s the impact of these new grid support functions, so allowing things like low voltage ride-through, volt/var support, on actually the anti-islanding conditions, and two, what about multiple inverters and multiple PCCs, and what’s their effect on anti-islanding? So the outputs of this – I’m not going to spend a huge amount of time This is probably an additional presentation that we could give in the future But the – really, the results showed that with grid support functions enabled, so volt/var control and frequency watt enabled, that the two second requirement was still able to be met It also showed that multiple PCCs did not cause trip times beyond two seconds, regardless of system topology So again, this was putting – scattering multiple inverters at multiple points of common coupling and then doing the same type of unintentional islanding test on that much larger system So one of the things that we would point out from this result is obviously these results are only valid for the inverters and designs that were tested, but they do show that the ones that – I would say popular designs that are out there do meet these types of requirements when looking at multi-inverters One of the big challenges we have with addressing this is the fact that it’s going to probably be impossible to test all the different types of deployments in all scenarios under all conditions And so one of the things that you really have to look at is kind of what’s the probability of these island conditions actually occurring and happening So as I move on in the presentation, we’ll discuss the probability of unintentional islands So this is something that – it’s obviously a discussion topic that’s been around for a number of years, and there was actually a study done out of the IEA – this IEA PV Power Systems Task 5 – that looked to benchmark probability of islanding, because the fact is, no matter what kind of distributed generation you potentially have, to create an island that sticks around and sustains itself, you have to have a very closely matched real and reactive power balance, and over time, that power balance can’t change So if all of a sudden you have a lot of load and a small distributed generation, that distributed generation isn’t going to be able to maintain an island When you separate into an unintentional island system, if you have a small amount of distributed generation, you most likely won’t be able to maintain an island condition And if you have an over-abundance of distributed generation, again, it’ll be difficult to maintain that island without moving outside the voltage and frequency nominal set points So the outputs of this particular study basically kind of said that there’s a benchmark risk that already exists for network operators and customers that’s on the order of 10 to the minus 6 for any individual person And then looking at specifically the risk of electrical shock associated with islands, under worst case PV penetration scenarios, was typically less than 10 to the negative 9 per year, so a number of orders of magnitude difference, because you have to maintain this massed real and reactive power case So in general, the outputs of this, and you can read more about it in the reference that’s on here, but balanced conditions are very rarely – I would say they’re very rarely

occurring, even at low, medium, or high penetration levels, and the probability that these balanced conditions are present and that the network can sustain them and disconnect at that specific time is virtually zero So there has been some work there that’s really identified what the probability of could be addressed Again, this is an area that you could do further research in I think we have a much better understanding of distributed generation than at the time of this particular report But it’s one of the things I think we have to look at as an industry, is really what’s the probability of this happening, and are we – are we kind of jumping on a specific area here that may have a really low probability of happening So in addition to this, there’s been a lot of work that kind of – to help utilities understand and assess the risk for unintentional islanding, and Sandia recently put out a report – it’s referenced on there – that kind of laid out a framework or guidelines that allows you to evaluate is unintentional islanding a significant risk or not And so I’ll walk through this Of course, we can talk about this a little bit in more detail in the question session But in general, what it did was it basically said unintentional islanding can be ruled out if you have these types of scenarios So one, where the aggregated AC rating of all the DG is less than some fraction of the minimum real power within the potential island, and not possible if you can balance the reactive power supply and demand within the island So again, this is looking at what the probability and possibilities of having a matched power both in real and reactive inside the island And then the other condition is this direct transfer trip or permissive power line carrier is used, you can pretty much rule out unintentional islanding would occur Then additionally, they said you should probably consider or do a little bit closer study if the potential island contains large capacitors and is tuned to a power factor which you could maintain the real and reactive power balance inside that island That’s one thing If you have a very large number of inverters, so if all of a sudden you’re looking at 100 or 1,000 inverters inside a particular area, that may warrant additional consideration Also, if inverters are from a variety of different manufacturers, oftentimes, you don’t really know exactly what type of anti-islanding technique they’re using in their systems Of course, you can ask the manufacturer to provide more information on that And then if you have potential islands that may include inverters and rotating synchronous generators, then you probably want to take a closer look at that, because the rotating generators potentially could trick the inverters to stay on longer than you would think But again, these are guidelines to help someone assess the actual probability of islanding So let’s move on here We’re getting close to the end And I talked a little bit about what’s the future of anti-islanding protection So here’s some kind of things that we can draw as generalized conclusions based on the presentation here So passive anti – or passive islanding techniques, or passive anti-islanding techniques, often have this non-detect zone So they definitely have some level of stable voltage and current or voltage and frequency which stays inside a normal area, and it’s hard to detect inside that specific space But it’s also hard for power systems to maintain that generation load balance for extended periods of time, depending on the control strategies of the distributed generation Active anti-islanding techniques are fast and work best on stiff grids So these are the techniques that are actively trying to move the grid around And we have definitely demonstrated that using the active anti-islanding, you can get a very fast trip time with these types of systems One of the challenges we see coming up in the future is that new integration requirements are opening up voltage and frequency trip points to enable grid stability at high DR penetrations So as you get more DR online, you may want to start opening up voltage and frequency trip points [inaudible] overall stability of the system So how does that mesh with unintentional islanding requirements? One of the things that we don’t have a lot of work on is multiple active anti-islanding techniques They may work with each other They may work against each other

It kind of depends on which techniques they’re using And as we look to future power systems, if we lose a lot of synchronous generation, they may not be as stiff So we may have – be forced in general to allow distributed resources to operate in a wider range of voltage and frequency settings So a couple of items for discussion I just wanted to throw these out, and I’m sure we’ll have some discussion here in a second But some comments on this Obviously, there’s a two second requirement in 1547 right now, and the question is, is this the right number? So some things to think about, it’s too slow for instantaneous reclosing, so if we have on the distribution system utilities using instantaneous or super-fast reclosing circuitry, that happens within two seconds, and so it’s not going to be coordinated with that anyway The other thing about two seconds is may actually be too fast for some communication-based methods, so some of the phasor based methods or other communication-based methods may require more than two seconds to determine if an island is occurring And the third point there is the need for active anti-islanding to achieve this number So two seconds is probably only going to be achieved with active anti-islanding And so these are things just to think about Obviously, we want to be able to detect unintentional islanding, but the requirement for speed there is something that we probably should address, as we see more and more distributed generation set points being opened up for larger voltage and frequency windows Okay Another thing for discussion is active anti-islanding, and is it needed I mean, we have a variety of techniques, passive anti-islanding techniques, communication-based techniques, but is active anti-islanding what’s really needed in the future? What happens when you have thousands of different techniques deployed on distributed resources? Will there be a way to test and evaluate the interaction of all these different techniques? Bullet point two there, should there be one single method that everyone must use? Well, in an active anti-islanding world, we actually tried this before everybody started patenting all their anti-islanding techniques Now there’s probably 20 different patented anti-islanding techniques out there, and it’s probably too late to get that genie back in the bottle, but there may be other permissive ways that – a single method everyone must use Finally, as we look at higher penetrations of distributed generation, will active anti-islanding work against maintaining grid stability? I mean, the way it work right now is it’s actively pushing voltage and frequency outside of normal set points So on very weak grids, that may be actually something that’s detrimental to the long term deployment of distributed energy technologies So those are just things for discussion I’m sure we’ll get into those here in a second Next couple of slides are just a set of references that can be used And then finally I’ll end right here, and we’ll start taking questions through the Q&A line So thank you, everyone So I think what we’re going to do now is if you have a question, you can submit it via the Q&A line or you can what they call raise your hand, use the raise your hand button, and we’ll put these up on the screen and then try to address them So I’m going to just read them out here In your opinion, when using the Sandia screening methodology, is it appropriate to move directly to requiring additional anti-islanding protection, direct transfer trip or other, once the load to generation test or other is failed, or should a detailed risk of islanding study be commissioned to determine the run-on times? Okay, so I guess I’m going to put this out there I mean, I think the Sandia screening methodology basically gives you a way to understand what are the potential risks I definitely would not say that once it gets – once you’re using that screen and you get to the load generation test, that you would immediately want to use direct transfer trip Direct transfer trip is an expensive option, and you have to look at how you’re going to pay for that, whether it’s utility provided or customer provided

It’s still going to be fairly expensive So I think that there’s probably – you just have to look at additional factors that may help you determine whether or not you need to require some form of embedded communications Again, as we went – discussed on those slides, some of those things look like the number of inverters that you have, do you have a mix of different inverter technologies, and get a better understanding of what kind of unintentional islanding or active anti-islanding techniques that those systems may be using I’d say based on the experience we’ve had to date, that inverters that use active anti-islanding do a very good job of detecting island conditions and disconnecting from the grid quickly I hope that answers that question Next – oh, okay What are the tradeoffs between trying to meet compliance for under-frequency load shedding and having frequency points that do not adversely impact a potential unintentional island? Well, obviously, this is something that really needs to be looked at It depends where you have your under-frequency load shedding set points Excuse me Now typically, those are going to be pretty low, 59.5 Hz or something It could be different, depending on your situation But you probably do not want your DG trip points to be tripping below your under-frequency load shedding trip points So you still want your DGs to be online If you have it the other way – let me put that – let’s put that [inaudible] I mean, one thing you have to evaluate is do you want to be shedding load or changing load before or after your DGs trip? And so I think that’s something that needs to definitely be coordinated, and is impactful, especially with how we’re opening up under-frequency load shed – or under-frequency trip points in the system Okay, 1547 requires coordination with reclosing Effectively, the islanding time limit is the lesser of two seconds and the reclosing time Agreed It does require coordination with reclosing You state that perfect load matching is unlikely, but even if the match is quite imperfect, the island may – the islanding time may exceed reclosing time, commonly used in New York Agreed So isn’t it misleading to focus solely on a two second islanding detection limit? I think that’s a valid point to bring up, is that you do need to coordinate with reclosing And this is a challenge we have with 1547 all the time, is just taking one of the requirements out of context with the entirety of the document So this person brings up a good point with being – having to coordinate with reclosing, but it also has to coordinate with the over and under-frequency trip points So there’s many pieces that have to be evaluated You can’t just look at the unintentional islanding two second requirement You need to effectively coordinate with over and under-voltage frequency trip points, the reclosing times, and the unintentional islanding I mean, I think at the end of the day, if you look at how unintentional islanding is set up in the document, it’s kind of like a catch-all that hopefully at the end of the day, what is what – the last resort to make sure that you’re not creating an island But you still would need to be coordinated, and you still need to be tripping outside or inside the trip points as listed in the document It may be one of the things that we – that 1547 evaluates, is do you really even need an unintentional islanding requirement if all your other things, like recloser coordination, voltage and frequency trip points, are needed? I’ll just throw it out there for discussion Okay. [inaudible] – >> Houtan Moaveni: Ben, some of the folks, is looks like they have problem to submit a question, so they send me the question via email If you don’t mind, please, allow me to read at least one or two questions for you >> Benjamin Kroposki: Sure >> Houtan Moaveni: So I received a question The question is asking like are active anti-islanding detection methodologies deployed on a per

phase basis? And I assume there is like a follow-up question, that do methodologies deployed for a single phase inverter, and the result of studies conducted on a single phase inverter, apply to their three-phase counterparts? >> Benjamin Kroposki: Okay For – obviously, for single phase inverters, the techniques are used on single phase For three phase inverters, I will say that the testing requirements require them to evaluate their ability on each phase So they are tested on each phase, so that you can’t just be doing anti-islanding on one phase of a three phase system So a three phase inverter, it is done on all three phases, and in order to pass the test, it’s tested on each phase individually I believe each phase individually and as a whole Does that answer that question? >> Houtan Moaveni: Yes, [inaudible] [Crosstalk] >> Benjamin Kroposki: Did you have another one? >> Houtan Moaveni: Yes There is another question Can conclusions from NREL’s recent work on multi-inverters, multi point of common coupling onto islanding with single phase inverters, be used in three phase applications? >> Benjamin Kroposki: I mean, I guess at some level you probably would want to conduct a very similar test on three phase systems But in general, the anti-islanding techniques are the same, whether they’re single phase or three phase And so therefore, I think you would have – if we didn’t do three – a three phase inverter, you may or may not trust that particular result I mean, I can’t say obviously without a doubt that that would be applicable to a three phase system But at the same time, the technique is basically the same, and I would – I would imagine that it could be extended to three phase >> Houtan Moaveni: Okay >> Benjamin Kroposki: Eventually, if we find there’s funding to do a three phase test, we could do that But I think in general, if you are logically thinking about the actual techniques that are being used, there should be no reason why a three phase system would give you different results at the end of the day >> Houtan Moaveni: Great I don’t have any other questions >> Benjamin Kroposki: Okay I’m going to move back to the ones that are listed on the board here It was mentioned active anti-islanding works best on a stiff grid What’s typically the short circuit values to determine a stiff grid? Okay, well, this is a good question that gets asked all the time I think we had some general guide – there’s general guidance for this in IEEE 1547.2, which is the application guide for 1547 It talks about grid stiffness and kind of when you can determine that I mean, I think at some level you can say a really stiff grid has got a short circuit ratio of – or I think it’s 25 That number sticks in my head But it doesn’t necessarily mean you have a weak grid at something less than that, but you’d probably have to evaluate it more closely There have been – there’s some older documentation that kind of says that’s what utilities were using as a measure for that But at the same time, grid stiffness hasn’t necessarily been evaluated with different islanding techniques The point here is that actually if you have active anti-islanding, you will move a weaker grid around quicker and trip probably quicker than a stiff grid, once you go into island – or it may not be able to work that well at all if you have a weak grid condition Okay? Next one, considering the footnote in 1547, islanding risk can be ignored if test was conducted, does this apply when you have multiple type inverters in a potential island where all of the individuals – individual equipment was tested to UL? So UL 41 and IEEE 1547.11 replicate basically the same test I guess my opinion on reading the footnote would be that if it – if you have a single [inaudible] system and it passes the test, then that should be considered anti-islanding or a non-islanding inverter If you – as we’ve shown, looking at multiple types of inverters in multiple places, that

most likely if you have multiples of these and they’re all UL certified, again, they’ve all gone through a pretty rigorous test, I would think that those should be able to pass So at the same time, you do have to evaluate that, and I think that this is one of the unknown places, as we see more and more of this type of system, what’s really the risk when you have a very large number of inverters connected together? Okay Next one You stated that the only inverter designs tested by NREL [inaudible] illustrate successful disconnection from the grid without requiring – within the required timeframes How can the general assumption that inverters utilizing active protection will successfully disconnect from the grid within the proper timeframe? >> Houtan Moaveni: Ben, can you please repeat that question one more time? >> Benjamin Kroposki: I’m trying to figure out what the question is here I did – so we stated that the testing that was done at NREL successfully showed that the ones that we tested disconnect within the required timeframes How can you – I guess the question is how can you make that general assumption I think what we’re making the general assumption is that if – I think if individual inverters meet the anti-islanding requirements, what we were able to show was that when you put multiples together, they still achieve and meet the disconnection requirements So obviously, we cannot test every single type of inverter, and every single configuration We just don’t have the resources to do that But we can show generally this would be the case Okay? Is there any testing done by NREL or Sandia that covers unintentional islanding where we have multiple single phase inverters and three phase inverters together? Okay, I think I answered that question together I don’t think we’ve done the mix of three phase and single phase together at this point But again, they use the same technique They use it on all phases So we would expect to see similar results Okay Two cases Distribution breaker opens and generator can stay connected longer, even though [inaudible] can result, is 1.73 per unit voltage on the [inaudible] phases for single line to ground fault can cause damage unless the DG provides ground source to prevent this I’m not sure what the question there is, but I would agree that if – you do need to take a look at the grounding for this type of condition and make sure that you’re properly grounded, so as not to induce the 1.73 per unit voltage on the system Okay? Next one, the Sandia report seems to specifically evaluate islanding conditions where the inverters are all using positive feedback-based anti-islanding Would this not make the final screening steps invalid for non-positive feedback-based anti-islanding? That’s a good question I don’t know – personally, I don’t know how many inverter manufacturers are not using the positive feedback anti-islanding techniques So that – I mean, it’s a valid question to have, and I don’t know the answer to it or how to answer it without doing some testing on some inverters that don’t have the positive feedback But as far as I know, everybody is using that Oh, maybe this is the second case Sorry, we got – okay, a delta Y station transformer and a large DG on the distribution side for fault on the transmission line The breakers on the terminal [inaudible] – I’m going to [inaudible] load, terminal will trip, and could leave a ground fault, effectively grounded systems, and you have to see [inaudible] Again, this basically is saying, which I would agree with, that you need to evaluate your grounding schemes for a variety of different type of breakers opening, and understand whether or not you have issues with that overvoltage condition Okay, the IEA study regarding anti-islanding risk considered a combination of rotating machines and inverters I do not think it had rotating machines It was solely PV inverters

Okay, next one Grid loads are not just resistors Typically loads typically have a substantial portion of motor load Will the conclusions of the NREL test on grid support functions be different if the loads used included substantial motor loads, causing a transient impedance [inaudible] from the inverter? Well, I believe when we were doing the testing, we were using balanced RLC load, so we weren’t using a motor load, but we were using the setup that’s normally done in 1547, which has both an inductive and capacitive element that is designed to maintain grid stability So it’s this resonant load tank that’s set to the Q factor of one, when we were doing the multi-inverter testing So we weren’t using a motor load We were using a resistive, inductive, and capacitive experiment Okay? Could you clarify whether or not midline reclosers should be considered in anti-islanding evaluations? I think one of the requirements for distributed generation in general is that they do need to be coordinated with reclosers I would – if that is a possibility of that opening, then I would definitely consider it when you’re looking at a potential island condition Okay, when inverter manufacturers do not disclose the anti-islanding methodology, how can one determine the specific type? And for your information, not all use the full positive feedback Okay Well, thank you for clarifying that I don’t – I don’t know how to force manufacturers to do anything personally, but I think that it’s something that manufacturers may want to make sure that they disclose in order to simplify interconnection to the grid, if they could let people know what kind of anti-islanding methodology they’re using Okay It’s a good idea to revisit and update the Sandia risk evaluation criteria with the new 1547 Agreed with that We talk with Sandia quite often, probably every week, not on this specific topic, but I think it’s something that will come up We’re doing some projects now actually out of what’s called the Grid Modernization Laboratory Consortium, of which both NREL and Sandia are members, and this is probably something that needs to be addressed into the future But I would agree with that Okay There is a concern related to potential overvoltage for single line to ground faults on tap sub-transmission systems with a delta connection on the high side of the distribution – oh, is – there is a concern? What is the typical solution used in industry thus far? The three VIO scheme relay and the sub-transmission system will require PTs at the high voltage side Yes, I would agree, PTs at the high voltage side is a very expensive proposition So I don’t know the answer to this question off the top of my head Maybe this is something that we take offline and try to do a little bit more in depth study on, because one of the big challenges you have is there are potentially solutions for overvoltages and for this unintentional islanding that require very expensive solutions So coming up with some lower cost solutions for addressing some of these issues probably would be critical to take a look at But we could probably reduce that to a subset here with taking a look at these overvoltages on the single line to ground fault, and what to do about direct transfer trip costs Okay What’s the – what is the minimum typical size of a project that justifies direct transfer trip? Like I said, I mean, I don’t know a specific number off the top of my head I see direct transfer trip used a lot on large distributed generation facilities, typically that have synchronous generators installed with them And this is – you know, we’re talking – the problem is that DTT probably costs on the order of $100,000.00 minimum, and may run up to $1 million, depending on what other kind of communications infrastructure you have And so for a large system, say 50 megawatts, 25 megawatts, 15 megawatts, you may have that in the budget to do For smaller scale, especially inverter-based systems, we don’t see this as a big requirement, mostly because they have the active anti-islanding techniques built into them, and it’s really

hard to get them to form an island condition So I don’t know if I can come up with a specific thing, but if I were to guess, it would probably be like 25 megawatts Anything lower, it’s going to be tough to justify the cost of DTT What’s the new voltage and frequency threshold – would the new voltage and frequency thresholds invalidate the results of the recent report findings? I don’t think it would invalidate the report findings The report findings are what they are, and we probably state what kind of voltage and frequency thresholds are being used in there If you start to require larger operational windows for voltage and frequency, we’d definitely need to probably go back and evaluate some of these things But you may also be evaluating whether you allow longer island run-on times Let’s just say if you had a ten second run-on time, then would that moot – make all these voltage and frequency thresholds moot? They could be as open as they need to be, but you could still detect and cease to energize within a longer time period The two seconds matches very well with the old requirements Let’s put it that way Okay? It was indicated that DTT with a communication [inaudible] should be expensive for a small project It was mentioned 20 megawatts Okay Maybe I just mentioned that also Okay? In your opinion, how much reactive power mismatch can rule out islanding? Sandia screened three and four even if reactive power match is very low It oftentimes depends on what kind of distributed generation device you have there In a lot of cases, if it’s a synchronous generator, and this actually ends up being for inverter-based also, you have to look at what the reactive power capability and settings are inside the island So it’s tough to make a determination here without knowing that typically, you know, these devices are set to run at a – at or close to a fixed power factor, whether it’s unity power factor or .9 power factor, lag or lead, so that it depends really on what those DR settings are If it’s allowed to move those around and provide reactive power, then that’s probably going to be the thing that either allows is to potentially support or not support an island Now if it’s fixed within a very tight range, then you probably will never be able to support the island, because you’ll never have enough reactive power inside there But again, it depends on what kind of loads are in the system, and may require additional evaluations Okay? Of the 20 types of active anti-islanding algorithms, can you assume each manufacturer utilizes these in the same way? For example, frequency shift, can you assume the same for each of the inverter manufacturers in a – I’d say unfortunately, no You don’t know what they’re doing, without asking them Obviously, things like Sandia voltage shift and Sandia frequency shift are out in the public domain, and are used I think some of the other ones that have a lot of patenting around them just are faster acting, if you will But as far as I know, pretty much if you implement all of these, that you will still be able to meet the requirements The problem ends up being is that you can’t really guarantee or assume you know what each of the manufacturers are doing The only thing that you can do is if they meet 1547 or UL 1741, that they’ve gone through a fairly rigorous test that’s really trying to trick the inverter into an island condition Okay? Next, can you elaborate on why you believe 20 megawatts being the threshold makes sense for DTT? Mostly it comes down to a cost perspective I mean, I think if you could come up with a communication-based method, maybe power line carrier permissive or some other thing, that was minimal cost for distributed generation to implement, that would also work just fine It’s not that I’m disagreeing with any kind of communication-based platform It’s just probably the overall cost We haven’t seen costs of direct transfer trip, mostly because they a lot of times require

a fiber or a leased line or something, to be able to make that cost competitive And it probably is a factor for what the actual installations cost on these projects But you have to think not only of a variety of sizes, say 20 megawatts, 5 megawatts, 1 megawatt, all the way down to are you requiring direct transfer trip on every 3 kilowatt PV system that gets installed? It would probably be better to have a consistent method that addresses people’s concerns, and you’re probably not going to run DTT to all those different types of examples So it’s something that the industry desperately needs, is a low cost method to ensure unintentional islanding requirements are met Okay? Do you have any examples of utilities that are testing or have experienced this – experience with this? It would be helpful as a reference point Every utility in the United States is having this exact conversation So I know we’re dealing mostly with the New York utilities and people on this particular call, but California is having this issue, Hawaii is having this issue Hawaii is having a trickier time in that their grid is a little less stable in terms of fixed voltage and frequency than the mainland grid So they are already – they already open up voltage and frequency set points just to stay online on that system, and now dealing with the unintentional islanding requirement has basically gone to a – the back burner, when they’re focused on just making sure all the DGs can stay online and support the grid in that particular scenario So we’re seeing this be an issue with practically anybody and everybody that is installing PV, and that happens to be practically every utility in the United States But even in the Southeast, we’re dealing with this issue Duke Energy, Dominion, Virginia Power, all of these utilities have exactly the same questions and are dealing with them differently Now experience with this Like I mentioned, the people that are I would say at the – at the leading edge or bleeding edge of integrating high levels of distributed generation, I would consider HECO to be one of these utilities They right now are not really worrying about unintentional islanding, and are much more focused on allowing all these distributed generation devices to provide voltage and frequency ride through and support to grid So it’s just something to think about into the future How do we transition to where we have a lot of this stuff embedded into the system? >> Houtan Moaveni: Ben, I have another question >> Benjamin Kroposki: Okay >> Houtan Moaveni: How effective is the PLC solution versus DTT from a technical perspective? >> Benjamin Kroposki: Well, from a technical perspective, we’ve I think demonstrated that the PLC solution works We’ve done enough studies and things on this I think it’s, again, a matter of cost You have to – you have to add it to your grid, if you will, the signal that goes out to all the distributed generation devices So you have to wrap that PLC signal around any distribution transformers or any things – you have to make sure that that can flow all the way But I think it’s definitely a valid signal out there, and option The challenges that I’ve heard that people have with it, potentially it’s something that would have to be embedded in distributed generation, so it’s adding a little bit of cost, but obviously probably much less than a dedicated fiber line And you’d have to make sure that the utility really is able to provide that signal without causing undo distributed generation trips, because now you all – everything has a very permissive signal to be connected, and so potentially you may have some issues with – with the ability to massively trip everything off simultaneously So there are some things that people need to think about Okay? Any other questions? >> Houtan Moaveni: No, nothing on my side >> Benjamin Kroposki: Okay Well, we’ve hit the end of the questions on our side, on the Q&A panel here

>> Houtan Moaveni: Great >> Benjamin Kroposki: I think that you mentioned that these slides will be posted somewhere, or sent out one way or another? >> Houtan Moaveni: Yes First of all, let me just thank you for a very informative presentation, and I would like to thank everybody on the call for joining us As you mentioned, the presentation and the recording will be posted on the Interconnection Technical Working Group webpage, and we will try to answer like some of the questions that were not answered during the webinar With that, I think this is it I would like to thank you one more time, and everybody, for joining us today >> Benjamin Kroposki: Okay And I’d like to thank everyone for attending Thank you very much Talk to you later >> Houtan Moaveni: Thank you Bye