All Episodes

[Episode #27] – Better Grid Modeling


Although it’s clear enough that energy transition is necessary and reasonable, and although we know that transition is mainly happening on the grid at first, there is still much uncertainty about exactly where on the grid different strategies can be tried, how much they can accomplish, and what they’ll cost, relative to the alternatives….not to mention how the rest of the grid will respond as different measures—like storage, demand response, rooftop solar, controlled dispatch, and so on—are implemented. What’s needed to answer all these difficult questions? Better models, including serious math, by serious researchers.

Fortunately, one of those researchers is willing and able to explain several years of her work in grid modeling at NREL and elsewhere. So tune in and put on your thinking caps, because this episode (Geek Rating 10!) is not for the faint of heart.

Guest: Marissa Hummon is a senior energy scientist with Tendril, a provider of customer-facing software to the energy industry, based in Boulder, Colorado. Previously, she spent five years at the National Renewable Energy Laboratory in the Energy Analysis group. She earned her BA in Physics from Colorado College and her PhD in Applied Physics from Harvard University. Marissa started her career in grid integration of renewables by looking at some of the core problems with modeling the intermittency and variability of renewable technologies. Before joining Tendril she worked on quantifying the value of demand response and storage technologies in wholesale electricity markets. At Tendril she is leading the development of a residential demand response product that balances the home owner’s comfort and the utilities’ production costs.

On Twitter: @Tendril

On the Web: LinkedIn Profile for Marissa Hummon

Recording date: September 13, 2016

Air date: October 5, 2016

Geek rating: 10

Chris Nelder: So let's bring her into the conversation now welcome Marissa to the Energy Transition Show.

Marissa Hummon: Thanks, Chris, for having me.

Chris Nelder: You've been involved with some very cutting edge research at the core of energy transition such as modeling the cost of renewable generation of high penetrations, modeling solar irradiance down to one minute resolutions, all sorts of modeling of demand response and electricity dispatch and now you're involved in developing customer facing systems that will deliver residential demand response among other things. It's pretty technical stuff and frankly a lot of it's over my head and I'm afraid will probably only have time to really scratch the surface of some of that in this interview. But I will link to the papers and so on in the show notes. So let's start with some of the research you were involved in at NREL, OK?

Marissa Hummon: Sounds good.

Chris Nelder: OK so let's start with an easy one. In 2012 you led a study that looked at the economic benefits of various compass orientations of solar PV systems which found that in certain locations under tariffs that paid more for peak hour generation. The economic value of the system was greater if the system were oriented a bit more to the west of due south. And I recall a whole slew of articles being published in the press saying you know basically, "you're doing it wrong!" and claiming that all PV systems should be oriented to the west and not to the south as has been the long established principle of solar PV design here in the northern hemisphere. So as a former PV designer of systems that coverage drove me nuts because it was clearly about rate design not just orientation and that other basic factors like latitude and shading also need to be taken into account when designing a system and further I knew that the generation of power would be greater over the course of a year for systems that were oriented toward the south. So it was really a tradeoff at some level between maximizing solar generation and maximizing income. What was your reaction to the way that report was received in the press?

Marissa Hummon: Well I think the paper and people's reactions to it get to the core of the issue. There is clearly a disconnect between the rate payer which we started recognizing as an actual human and power generators and PV owners were actually paid the market rate for their power production, they would in some regions benefit by orienting their systems slightly to the West. Except in Florida where the cost of energy peaks in the mornings so it should actually be oriented slightly to the east. But if I were a utility and I was looking at installing you know a ten megawatt system of fixed orientation, I think the prudent thing to do would be to try to align solar profile to the load profile and then try to claim some capacity credit for it. I think in California, this is a pretty crucial since misalignment of that load and solar is causing this really large early evening ramp and...

Chris Nelder: The so-called duck curve.

Marissa Hummon: Yes exactly. And that's costing California a lot of money to meet because essentially they turn on a whole bunch of power plants to meet that rapid rise in that load profile.

Chris Nelder: Right.

Marissa Hummon: Alternatively you could couple the with storage and suddenly due south is clearly the best orientation because you can maximize the power generation from solar and store it and use it at the right time of day.

Chris Nelder: Yeah. OK. Now three years ago you were part of the team that looked at various ways to value grid power services that storage can provide such as energy arbitrage, power regulation services, and contingency reserves. These are a few topics that we touched on in one of the earlier episodes of this podcast on storage. You modeled these for both vertically integrated utilities and for a market environment. Can you summarize some of the key findings of that study?

Marissa Hummon: Sure. So that was a really big study and it went back to the basics of how do you approach measuring the value of storage? The core of that study was to look at how much value was being left on the table when you simply took the actual past prices and looked at the energy arbitrage opportunity. The biggest takeaway was that when you basically use a price taker model versus using an economic dispatch of storage you leave about 30 percent of the value on the table. And so we did that because we wanted to make sure that when people are making a choice between storage or say a natural gas plant that they didn't underestimate the value of that storage in an actual operation of the system.

Chris Nelder: Right.

Marissa Hummon: So the second probably major point that that paper made was that by providing both energy and reserves you could also increase the value of storage. It's important to model that in a true economic dispatch sense because reserves markets tend to be very shallow and if you overwhelmed that market with storage you could quickly collapse it basically bringing the price down to zero. So that was probably our first foray into really understanding how to model something that had the potential to bid in at zero marginal cost because there's no fuel cost for a storage device. And that modeling it in an economic dispatch situation allowed us to kind of see the breaking points of the system when you added storage to it.

Chris Nelder: So did it appear that the storage services were being properly valued at that point? And are they being valued any more accurately now three years later? And if not what sorts of regulatory changes do we need to make to value storage properly?

Marissa Hummon: Well I think that it might be a question about whether or not we're valuing storage properly or it might be a question of whether or not systems are able to dispatch it properly.

Chris Nelder: Great point.

Marissa Hummon: So in the last couple of years a lot of states, probably about 10, have made an effort to grease the skids for storage. They've established financial incentives for both utility scale and distributed storage facilities. They've initiated some technical potential studies that allow for new policies to be put into place but probably the biggest change has come in California in the last couple of weeks where California's Independent System Operator CAISO adopted new rules about how storage is actually bid into the market. So up until very recently when storage was bid into CAISO, it was bid in at a fixed state of charge, I think of 50 percent and then it was dispatched from there and now storage can bid in its actual state of charge. And CAISO will basically know that state of charge and then dispatch that charge discharge cycling accordingly. So I would say going back to the study that we did in 2012 or 2013 at NREL, what we did is we modeled storage the way you would expect a system operator to want to use storage. In other words put storage on the same footing as a generator. If you model all of the properties of a storage device with the same level of fidelity as you modeled the properties of a generator in terms of... You know, when you model a generator you have its heat rate curve and its minimum load point and maximum capacity and you know it's on-off cycling time. And if you model those same characteristics of storage then you can co-dispatch them in a way that truly recognizes the value that storage brings to the system. So instead of using some heuristic about oh well let's charge that storage device when prices approach a low point and discharge it when prices approach a high point, now you can you can choose to discharge that battery to avoid starting up a generator at a high cost point in time and you could avoid shutting down a generator at another point in time by using the excess power from that generator to charge the storage device. So true co-optimization of a storage device and generation captures that full value and that's what I think has happened in the last... basically in the last month in California is they've really changed the way storage is able to bid into the market and therefore change the way storage is going to operate in the market. So I expect that maybe over the next year or so that we'll start to see storage basically gain more and more traction in the California market because it's getting paid closer to what value it can actually offer to the system.

Chris Nelder: One of many examples it seems that you've been involved and where just simply providing more discreet and comprehensive data about the operation of the system components is really just a key hurdle that we have to overcome.

Marissa Hummon: You know I think basically when it comes to operating the grid it's always been operating generators against a transient load.

Chris Nelder: Right.

Marissa Hummon: And with storage or demand response coming onto the grid we've taken baby steps to do it by just allowing them to operate kind of at will but their true co-optimization with the rest of the grid is what's going to allow them to operate at their maximum value to the grid.

Chris Nelder: Well I think what most people really want to know is is storage worth it? I mean can we say at this point what the relative costs are of integrating more storage into grid power systems versus just simply adding more capacity using natural gas turbines?

Marissa Hummon: Well I'm not sure that asking the question "is storage worth it?" if that's the right way to frame it I think this question should be pre-suppose by the need for the grid to evolve to be sustainably operated with significant penetrations of renewable resources. I clearly spent six years at the National Renewable Energy Laboratory and so I'm pretty much in the camp that if we don't evolve the grid to be to be operated primarily on renewable resources like solar and wind that we can decarbonize the whole industry. And so is storage worth it? Absolutely. Now what kind of storage? I think that's definitely up for debate. So battery storage is pretty darn expensive, but take for instance electric vehicles. In the last year I've seen a distinct shift amongst OEMs away from the stance that there is no way that we should ever use batteries for vehicle to grid support, to looking at vehicle to grid support as both a source of revenue from them from a capacity standpoint but also as a viable option because they don't think it's going to degrade the battery any faster than normal driving operation. So I think there are lots of potential sources for storage that don't involve you know stationary lead acid batteries in a warehouse or something like that.

Chris Nelder: Sure. Absolutely. And you know I constantly try to remind my listeners that there's really a whole lot of different storage technologies out there besides the conventional batteries they're familiar with or even the EV batteries. You know there's the box of rocks on rails approach that tried out there in Nevada which I think is really interesting. There's all sorts of thermal storage devices that most people don't even know about like you know a vessel full of hot gravel kind of a thing. There's compressed air energy storage, there's all sorts of things that I think the research is really just sort of getting started. But your point is well-taken that it's really much more about as you say as a co-dispatching and the optimization and you know what are the real opportunities to capture grid benefits depending on how you integrate it?

Marissa Hummon: Yeah I agree. I think that that storage is definitely going to be part of the future of the grid. You know the form is that from a technology standpoint it is up in the air and probably will end up being a mix of technologies. But fundamentally the operation of the grid has to evolve and that's like really at the core mathematic sense of the operation of the grid has to evolve to be able to take into account the characteristics and constraints of all those different kinds of storage devices.

Chris Nelder: Now you contributed to another 2013 study which looked at the value of concentrating solar power systems otherwise known as CSP which are equipped with thermal storage. And I will admit I've been really frustrated at the lack of utility scale solar plants that are equipped with storage. I mean it just seems like such a no brainer way to get around the intermittency issue. But we didn't build them with storage at first which I think was mainly because the RFPs that we put out for utility scale solar didn't require it and then CSP just started losing out to solar PV on cost in the market. Those solar PV plants didn't have a storage capability either. So what do you think about the value to the grid of these CSP plants equipped with thermal storage?

Marissa Hummon: So when we talk about value I want to clarify that we talked about the operational value and so from a technology bidding standpoint PV and CSP with thermal energy storage both bid in at zero marginal cost right? There's very small variable operating costs and there's there's no fuel cost to it. So clearly when you bid CSP with thermal energy storage into the market it has a higher value proposition than PV because it has that story component. It can basically be the completely dispatchable resource. It's so to speak the holy grail of renewable technologies. It's completely dispatchable, comes at zero marginal cost and has no carbon impact on the grid. But those systems are huge and expensive. There are tons of steel and glass and in maybe 10 years maybe the cost of that will come down to meet the capacity cost curves but at the moment, solar or PV without thermal energy storage or without any sort of storage beats CSP with thermal storage out in the marketplace. Now I think that if we have a carbon tax and had a way of maybe properly valuing the capacity credit of CSP with thermal energy storage versus PV then we can start to see better competition between them. But it's going to be another 10 years or so.

Chris Nelder: So you really think we're just sort of waiting on policy to make thermal storage with CSP viable again?

Marissa Hummon: Yeah I think so. And although I will say I think an alternative to shoring up individual plants or individual solar generation sources with storage is to broaden the balancing area, put in high voltage transmission lines. So spread it out the transient dynamics of solar generation over a wider region so you're not getting so much variability in just one spot. So I think there are probably competing options and places like NREL and other national laboratories are actually really good at evaluating whether or not we should build a high voltage transmission line or you know add storage to the grid.

Chris Nelder: So do you think the value equation or maybe the economic viability of storage equipped CSP plants is any different now than when you did that research?

Marissa Hummon: I say not really. By some measures I actually think that market is a little bit flatter than it was even four years ago. We often look at high price natural gas scenarios in order to see a realistic value of storage or an economically viable value store.

Chris Nelder: And now nobody thinks natural gas is ever going to be expensive again.

Marissa Hummon: Right. So that makes it a little hard to see that there's going to be a lot of headway on storage at least in the near-term future unless we change some rules about firm capacity or the dispatchability of a resource.

Chris Nelder: Yeah and I think that's really a key point here is that we just really have not really offered the full suite of value streams to these kinds of plants that are already available now to say a natural gas peaker plant.

Marissa Hummon: That's right we don't. And I think I think that might change but as I said I think there are competing options for the way the grid could evolve and I think, not to change the subject of the interview. but to talk about demand response... That's a source of readily available storage capacity located at the same location of load and might end up having a higher value proposition than you know CSP with thermal energy storage or a large battery bank you know somewhere out in the middle of the desert.

Chris Nelder: Well all right since you brought it up let's talk about demand response. You've done a lot of stuff that's focused on demand response. You contributed to a report that presented an approach for predicting how various kinds of loads might participate in demand response markets by providing three ancillary services, energy and also capacity down to an hourly resolution for the entire year which is quite a feat. Then you use that information to model how these demand response services might actually be integrated into a production cost model. Now this is pretty deep and technical stuff and we probably can't do it justice in this interview but can you briefly summarize that research and explain what it can tell us about the potential for demand response.

Marissa Hummon: Sure. I think I think I'd like to start by saying that demand response really is hinged on three main points. First we have to find and quantify the flexibility of loads. In my experience in looking at the residential load sector really closely there is a lot of fat in the system. In other words there is a lot of energy that's expended on a load that doesn't actually do any work so to speak right? It's not that lighting a particular task that actually was needed to be done at that time or you know it's washing a load of dishes at a time that wasn't crucial to the operation of the rest of the household. There's a lot of fat so to speak. The second thing is that you know once we find and quantify those flexible loads, we need to be able to figure out how to cost effectively equip them with controls and communication technologies. I don't think we can rely on the human or even you know a building operator to be in charge of managing that load on a moment to moment basis. It needs to be something that's part of that you know "smart grid network". It needs to be easy. It needs to be intelligent and effective. And then lastly I think to get demand response right we've got to find the right payment or incentive structure that offsets whatever opportunity cost we might have had. So even if there is a lot of fat in the system the person still is probably losing out on some sort of opportunity or the system or the business is losing out on an opportunity and they need to be compensated for that. They need to figure out a way to...

Chris Nelder: To get paid for opportunity cost.

Marissa Hummon: Yeah. So when I was at NREL we really tackled the first of those issues. Finding and quantifying load flexibility. Figuring out how much it was worth. And it was actually a surprisingly hard task because essentially you took a bunch of people who were really good at economic dispatch. They had been operating production cost models, quantifying the effect of various constraints on generators. You know we get a lot of studies where we would like swap out coal plants for natural gas plants or you know put PV systems into a grid and kind of you know see how the system would operate. But we had really basically left load as a fixed quantity that had to be met on a five minute basis so running real time models every five minutes. And now suddenly load was no longer a fixed quantity in the model and needed to be dynamic. And so given the tools we had we said okay let's make virtual power plants out of these demand response technologies and figure out how they would operate in conjunction with the rest of the grid. And that goes back to that original point I made which is if you want storage or demand response or you know any of these emerging technologies to be part of the operation of the grid. It had to sit in the same mathematical space as the rest of the generators on the system. Otherwise a system operator wouldn't know what to do with it. And so we did something at NREL with our colleagues over at Lawrence Berkeley National Laboratory that thing really had ever been done. We sat people who knew load really really well in the same room as people who knew the system really really well and painstakingly went through and quantified what were the characteristics of load that allowed them to be operated by a system operator. So how much flexibility is there in cooling a house by two degrees earlier in the morning and how much lead will that shift in the afternoon? And you know do we have four hours of load flexibility or six hours of flexibility? And we basically went through and did that for 13 end uses across the demand response spectrum. So everything from industrial out to residential loads and basically got them formulated into something that could be dispatched and that basically allowed us then to ask a whole bunch of really interesting questions. Things like well should a utility in Southern California invest in agricultural pumping for demand response or should they invest in residential cooling loads? Should we encourage control technology companies to take their control technology all the way down to being able to operate at four seconds, basically mimicking something that would run in a regulation market or is you know is five minute or 15 minute load dispatch going to capture the most value for them? And so that was a big technological breakthrough that really opened the door to being able to answer some of those questions. The work now that I'm doing it Tendril takes on the second or third issues the you know how do we equip loads with control and communication technologies and how do we find that balance of opportunity cost and payment? And those are pretty hard issues. We're definitely working on them in conjunction with utilities. Figuring out how much fat really is in the system and whether or not customers are are willing to basically you know let somebody else play with loads on their property in order to gain a few dollars.

Chris Nelder: Yeah. So we should probably explain for the listeners a little bit that that Tendril, where you went to after your research at NREL, is a provider of customer facing software to the energy industry which is based right here in Boulder Colorado and you've been working there on optimization and control of residential energy systems which is essentially a continued focus on demand response just in the residential sector. So what possessed you to take that leap and what appeals to you about tackling residential demand response? I mean particularly considering that most demand response today I think comes from the commercial and industrial sectors.

Marissa Hummon: I'll say I came to Tendril because I wanted to build something that would really impact the energy sector. At NREL my research was 10 to 30 years out and it was fascinating and I got to think about really big problems but I also felt like I was a little too disconnected from the real world to know how would we actually get a utility to change the way they operated the system?

Chris Nelder: Yeah.

Marissa Hummon: So I came to Tendril with this viewpoint that if we could build something that kind of sat in that niche basically as an intermediate product between where we're currently at which is we meet load on a minute to minute basis and we don't expect anything out of load to the ideal version which is load a fully integrated part of grid dispatch... I feel like we would make a big leap forward. And so that's what I came to Tendril to do. Tendril's fantastic place to work with smart people and great problems and great leadership. And I think the thing that I didn't know before I got to Tendril that has been a huge blessing is that small companies like Tendril actually influence regulatory filings. We sit down with program managers at utilities and help them write the regulatory filings. We can make a huge impact on how utility executives evolve their companies and that is a tremendous amount of influence and is pretty darn amazing.

Chris Nelder: That's something the national labs don't do.

Marissa Hummon: Well and they don't do it for a reason. They can't. It would be really awkward for... and probably a little inappropriate for government laboratory researchers to write regulatory filings. That's really outside their their space. So I think that small companies like Tendril are going to change the way utilities operate, the way the electricity system operates.

Chris Nelder: So you really see significant potential in the residential demand response sector.

Marissa Hummon: I will say when I interviewed at Tendril I kinda asked them, "Are you sure that residential demands is the thing you want to do? You know there's not a lot of value in it. Let me show you this paper I wrote." And they were like, "Well, but that's our space." And I said, "OK well I'll come check it out." And so the first thing I did when I came to Tendril is I set up a small field test so I took connected thermostats, put them in a dozen Tendril homes and I started playing with some thermal models that Tendril had built and started trying to quantify how much load you could shift away from peak. And we did this last summer and I was blown away. I did not realize how much untapped thermal mass there is in a house and particularly in some places like Colorado where we have overnight low temperatures and high daytime temperatures there is a tremendous amount of efficiency opportunity and peak load reduction opportunity just by more intelligently scheduling cooling operation. So if you can run your compressor when the outdoor temperature is 10 degrees cooler you can boost your efficiency by a point or two and then you can basically precool the house and and coast through an afternoon peak without the residents of those houses really knowing what's going on. And I can tell you they're occupied during the day. I messed up a few times. I made people a little bit too cool in the morning or a little too hot in the afternoon and I definitely got some calls but it was... I was actually really surprised about how much potential there is for basically using the house as a thermal mass, using it for thermal energy storage.

Chris Nelder: Well you know I've often wondered how well demand response and load shifting really works for air conditioning. I mean it's not hard to imagine shifting the operation of a hot water heater or refrigerator. For example the off peak hours you know that's something you can do without customers really noticing it but AC seems to be another matter. I mean people are actually pretty sensitive to just a couple of degrees change in the temperature. I mean how possible is it to shift the cooling loads or are you just making people hot in the afternoon?

Marissa Hummon: I would say we were actually doing a pretty good job of not making people hot in the afternoon. In order for this to work you do have to make them a little cooler in the morning. And some people are more sensitive to that than others. But about 75 percent of the people that we tried this on are fine dropping you know two to three degrees in the morning and then not really ever reaching higher than their their normal afternoon peak temperaturein the afternoon. And so I think that there's a pretty large potential for that. And then you really got to couple that with the incentive payment right? So take somebody in California who's on a time of use rate, like Southern California Edison. Their peak kind of use rate is 46 cents and their off peak period is 11 cents. That's a lot of money on the table if you can shift load from one time period to another. And I mean we've calculated for people on that kind of use rate that if you shift you know 50 or 100 percent of your cooling load out of that 46 cent window you can save anywhere from three to five dollars every day. And that's that's real money. That's not one extra coffee a month that's more like a Starbucks coffee everyday.

Chris Nelder: Yeah. OK well what kind of incentives are really needed to get residential customers to participate in demand response programs? I mean not everybody has a 4x differential between on peak and off peak rates right?

Marissa Hummon: That's right.

Chris Nelder: And since not too many residential customers have to pay a demand charge I assume the time of use rates is in fact the main tool.

Marissa Hummon: So I think the time of use rates are a transitionary tool. So I think that at the moment time of use rates are necessary because that's the way our utilities are set up and that's the way regulations are allowing utilities to pay for that sort of service. But I think in the long run just the way the cell phone industry went away from you know paying on a minute by minute basis and paying for peak minutes and off peak minutes I think in the long run customers are going to demand that utilities will evolve to have customers on a fixed rate. They're on a fixed rate but they've coupled that with being able to do some load management that minimizes the cost of the utility providing that energy. So you're right a lot of places don't have a time of use rate with a 4x differential but every utility experience is more like a 10x differential on peak load days during the summer. And so they're essentially riding through that cost by making a profit on providing energy during off peak times, but in the long run I think if we can have better control over that load from a daily load shifting standpoint, then we can move people onto a guaranteed fixed bill, a flat $100 a month for your energy and allow the utility to have the tools to manage their costs against that. Does that make sense?

Chris Nelder: Yeah but it surprises me because it's actually kind of the opposite of what I thought you were going to say. You know I thought what you're going to say was what we need to do is match the prices that customers experience more closely to the shifting costs that the utility experiences over the course of the day. And that was going to lead me into some thoughts about you know something we discussed on this show before, the potential for highly variable or nearly real time rate structures. But instead you've gone the other way and said we just need to raise the base price to a high enough level that utilities can just ride through the whole thing.

Marissa Hummon: Well and I would assume you were suggesting that we should raise the base price. I'm suggesting that instead of putting the burden on the end user, the customer, to manage their loads against a time varying price signal, allow the utility who's really good at understanding the economic dispatch of the system to manage both generation and load. And basically in return for allowing the utility to manage some parts of your load they're going to actually give you a guaranteed fixed bill that's lower than what you're normally paying. So we've basically been working on mathematics around cooling loads and we've found that when you manage cooling loads properly by taking into account you know everything from occupancy to the weather outside to the normal pattern of operation to also a little bit of behaviorial energy management around you know what you actually don't notice when I bump the temperature up by one degree... That we can reduce total cooling loads by about 20 percent. We can strategically reduce those during peak load periods and so take somebody's bill that was normally $120, reduce it to $100, make it guaranteed and then manage that load in conjunction with all the generation resources. So basically my stance is that the highest value proposition for load management sits with the system operator. They are the ones with the most information about how to co-optimize load and generation. So even if you gave the customer you know real time granular rates they will never make as good of a decision or as positive as of an impact on the system as if you gave that control or that decision making capability to the utility. And I think that the fundamental core of this though is that the utility or system operator has to have tools to do that well. So you cant manage somebody schooling system unconstrained right? The customer needs to be able to set some expectations and they do that not by filling out a form but more like you know this is my normal operating temperature range right? I like it at 72 overnight and when I'm gone during the day you know please don't kill my dogs so don't let it go above 78 degrees. And they do have normal interaction with their thermostat right? They set up a schedule or they come over and they adjust their thermostat and the system operator needs to be able to take in that information and make good choices that meet the constraints of the customer around their comfort and their expectations of what that load is going to do for them while co-optimizing that with the rest of the system. So does that make sense?

Chris Nelder: Yeah. No it really does. You know I'm just thinking here about some of the other ideas that we've toyed with on this show. I mean in Episode 20 with Eric Gimon we talked about you know he had this concept of like a little box that would negotiate with the utility on behalf of all the appliances in the house to do sort of real time transactions involving things like demand response or at least load shifting so that you could actually implement more of a variable rate structure or more of a real time rate structure but without the customer having to be involved in it because the negotiations and all that stuff would be done by this box. And you know I mean I've often wondered if real time rate structures or anything close to it is even going to work for the residential market because obviously customers have no familiarity with rates that vary that much or with managing any sort of a user interface to control their demand to respond to a rate like that. So you know I've often wondered like how granular can the rates be at the residential level and how granular do they need to be to make demand response work? And I think you've you've offered kind of an interesting answer to that.

Marissa Hummon: And I think that it depends on the implementation of that you know transactional framework. I'm pretty familiar with PNNL's work on that. I think that the core issue is is there an actual co-optimization happening at the operational level or is the operator going to issue a signal that basically calls for load or calls for a reduction in load because if it's the latter, I think you're always going to leave some value on the table. Just the same way that you know in that storage study that I talked about price taker models where basically storage gets a price and decides whether or not it wants to charge or discharge based on basically past history of prices right? Oh this looks like a low price. I think I should charge now. Oh this looks looks like a high price I'm and to discharge now. Or do you actually put the charge discharge decision making inside of the economic dispatch so that real time transactional network that has basically the house getting a signal and then the house makes some decisions... I think you still leave some amount of value on the table and that value is probably the things that aren't captured in the marginal cost of energy. So the marginal cost of energy only captures basically fuel costs and variable O and M costs costs and it doesn't capture all those start up costs. And as we move from you know a grid that has five to 15 percent renewables to a grid that has 50 to 80 percent renewables basically the vast majority of the costs are going to be in a start up and shut down costs because there is no marginal cost for solar or wind. And so the marginal cost framework is really going to fall away. It's not going to be the thing that allows us to make good economic dispatch decisions. You're going to have to make those decisions on the basis of these other costs like startup and shut down costs or basically you know transmission switching or other more complicated decisions. So yeah.

Chris Nelder: Well okay but let's kind of get back to the core question here. How granular do the rates really need to be? Especially for something like residential demand response.

Marissa Hummon: So it could be that there is no such thing as a granular rate. Maybe there is no per kilowatt hour rate at all for loads anymore? That we just say, "You're a house. You know, in the past you've used this much electricity and so were going to charge you a flat fee this covers demand charges and energy charges and transmission charges and if you greatly exceed our prediction of what your load is going to be next year we're probably going to bump you up. And if load management basically yields a much more efficient use of loads in the next year then we're going to move you down. Instead of $100 we're going to only charge $95 next year." I realize that has a lot of problems in terms of like transient populations and you need data on the household the people in the household in order to make those sorts of decisions, but I think it's a paradigm we should at least think about as opposed to this idea that we're just going to make the network you know nodal prices go all the way out to the individual house and have individual houses responsible for their part of the grid. Now there's a lot of costs in the grid that aren't captured by that marginal price. And so that granular rate is probably not the most prudent signal to use when doing load management.

Chris Nelder: Wow that's some outside the box thinking right there. I like it. You know, I mean it does sound to me though that this is going to impose a kind of a whole new kind of a burden on regulators.

Marissa Hummon: I think it will. I mean it fundamentally changes the face of the utility and it puts a tremendous pressure on the regulators to make sure that that utility is doing good by the customer.

Chris Nelder: Well not only that but you'd have to have some way of mitigating, I don't know, what amounts to basically a kind of moral hazard of customers that aren't necessarily paying for energy anymore.

Marissa Hummon: That's true. And I haven't even really gone this far in this thinking, but what happens when a customer puts solar on their roof? So some of the problems with solar at the edge of the grid are grid stability issues that we don't really capture at all right now in our rate structure. If you have solar you have smart inverters, do we lower your fixed bill a little bit more? You know is there any incentive to own your own solar yes we're going to basically you know co-opt that solar into the grid dispatch and then just change your annual electricity bill accordingly?

Chris Nelder: Yeah. Then it's no longer sort of an asset that you own that belongs to your house that directly impacts your bill. It's more of a model where the utility's basically renting the square footage of your roof.

Marissa Hummon: Yeah.

Chris Nelder: It's a different model.

Marissa Hummon: It is a very different model and I don't think it works for all end use loads. Yeah I don't expect Xcel Energy, that's my utility... I don't expect them to know about whether or not my cell phone is plugged in or if I'm watching television. I certainly would be pretty darn irritated if they told me I couldn't wash my dishes before I leave for work in the morning. But you know I think you could see a system where you're like OK the big loads in my house, my electric vehicle, my air conditioning system, my pool pump, my electric water heater... Those I will let Xcel Energy co-opt, use as best to serve the needs of the neighborhood and the grid. And in return I get basically an annual credit of you know 150 bucks or something like that. And you know I think probably the biggest problem with this is that you know we have a long ways to go in terms of educating you know the system operator all the way out to the end use customer on how this could work. And I think there are some other industries that are great models for this. I mentioned the cell phone industry but I think you know online streaming is another good industry to talk about. So you know Netflix doesn't curtail my movie watching on Saturday night just because everybody else is also watching a movie right? They have figured out how to handle that sort of capacity problem. They might slow the feed down a little bit in order to load balance but they don't charge me more for a movie on Saturday night nor do they cut me off. And I think the electricity industry needs to move that same direction. They need to find the places in the system that can provide an opportunity for load management and manage those loads and then let everything else kind of slide through, you know, like my dishwasher.

Chris Nelder: Yeah that's a really interesting analogy actually because how does Netflix do that? Well a lot of that is using the cloud. And so then you have to ask well what's what's the cloud equivalent for the utility? I guess it's a larger balancing area.

Marissa Hummon: I think it's a combination of larger balancing area, more more flexible transmission operations... So you can look at moving power over larger distances which is again a balancing area problem. Basically you know the reason why we've been able to meet load on a moment to moment basis is because we do have large enough balancing areas that loads don't look quite so flickery.

Chris Nelder: Chunky. Yeah.

Marissa Hummon: Chunky. You know if you look at a single house load when the air conditioner goes on you know it goes from 500 watts to 5 kilowatts in about 10 seconds. And so if everyone did that at the exact same time the system really actually couldn't deal with it. And so there is that sort of variability once you have those loads under being able to be you know communicated with and controlled you can really do that sort of you know basically broad storage over a whole region. So you're looking at when you think about pre-cooling a house you're really thinking about spreading out all of those cooling loads over a 12 hour window instead of trying to have them all condensed within a three hour window. There's tremendous flexibility once you get to that aggregate optimization level.

Chris Nelder: Yeah. As long as it works I mean I don't know if this is fair - probably isn't - but just now you brought to mind kind of the opening scene from Brazil know where the HVAC system goes haywire and the coffee maker turns on at the wrong time and you know the toast flies out and everything goes nuts because the central utility operating things is not working properly.

Marissa Hummon: There is definitely some risk of that. And I have some thoughts on how to do that from a mathematics standpoint. I haven't really tested any of them, but I think this might go back to a little bit of maybe what Eric was talking about with the real time transaction box and I think that you know you could imagine a scenario where some loads are basically under the direct control of the system operator and other ones are under advisory control. So I would definitely never put my coffee maker under the control of the system.

Chris Nelder: There are some things that you just don't want to risk, you know?

Marissa Hummon: That's right. But so take a an electric vehicle for instance. The way we charge the battery in a electric vehicle has an impact on the long term efficiency of that battery. And I don't know if it's reasonable or even necessary for the system operator to know about the detailed temperature across the interfaces of the battery but the charging system does know about that. It's basically that's the mechanism for keeping that battery from exploding. And so maybe we send a combination of direct load control signals and advisory signals and maybe those resources that only get advisory signals get a discounted capacity credit. Because they're not quite as dispatchable as something that's under direct load control. So we're now in the realm of like market design and a little bit further outside of what I've actually worked on that. But I think that I think all of these things are possible. We just need a lot of willing participants to field test how this would work.

Chris Nelder: Well in your work on demand response have you had a chance to look at EVs as demand response assets?

Marissa Hummon: A little bit. I did some work with the California zero emission vehicle program and specifically we looked at them in conjunction with a whole bunch of other demand response assets in California. I think that that is not a technical potential issue. There is tremendous potential there. It's really been the OEMs that have been more reluctant to allow those vehicles to even be part of you know field test operations to see how well they provide services to the grid, but you can see that the electric vehicle manufacturers are positioning themselves along that load management food chain. They are you know establishing strong relationships with their vehicle owners, they're supplying charging stations, they're managing the settlements of those charging transactions and I am pretty sure that as their penetrations increase, as the number of electric vehicles on the road increases, you can see them forming aggregation companies that allow them to bid that capacity into markets. And I think California will probably be the first place that that happens.

Chris Nelder: Yeah I agree. I actually wrote a report for RMI this year on using EVs as demand response assets, as a sort of energy resources.

Marissa Hummon: Cool.

Chris Nelder: Just by managing the charging, not even getting into the whole "V to G" thing.

Marissa Hummon: Oh okay. Yeah.

Chris Nelder: And it is pretty interesting, you know, once you start looking at how big the numbers can get it starts to look pretty significant. I mean it could actually be the largest flexible load on the system in time.

Marissa Hummon: Absolutely. And if it's not under some sort of control it could also be the thing that just brings down the grid.

Chris Nelder: Absolutely or at least causes your peak power pricing to really spike.

Marissa Hummon: Yes exactly.

Chris Nelder: So have you had a chance to actually quantify the total market potential in some way or another for residential demand response? I mean I don't know how you quantify that in terms of dollars or percentage of the load or whatever.

Marissa Hummon: Yes. So across the country, no. For regions, yes. I think in general when you look at cooling loads it is completely reasonable to expect that 50 percent of the cooling load is manageable on a day to day basis. I think that when you do that the problem is that once you get past about 1 percent of that load doing that you shift enough of the prices around that you'll start to decrease the value proposition of that. Right? So every market if loads were completely manageable and you could basically make them flat there would be no additional value in managing them. And so it's a hard question to answer.

Chris Nelder: Yes it is.

Marissa Hummon: But yeah I'll stick with that. I'd say pretty clearly we can manage about 50 percent of the cooling peak load out at the peak load window. And let's just say that that window's about four to six hours long.

Chris Nelder: Well that's pretty impressive. And what about the heating side?

Marissa Hummon: On the heating side... Well so we've done a couple of different things of looking at natural gas. So specifically, looking at heating and areas where there are natural gas transmission constraints or flow constraints so in the Northeast and in the L.A. basin in California both of those areas have basically a competition for gas flow to end use heating versus to natural gas peaker plants that provide electricity and during those same time periods in the winter. So when you do that you have to construct a price signal that doesn't exist right now for natural gas because that's not something that's generally time varying. And you can do some really cool heating load shifting around that. And I haven't quantified exactly the potential because it's on a... more like on a year to year basis you got to see what the actual overall system needs are. So I don't know what the long term value proposition is.

Chris Nelder: Well what about where heating is electrical heating rather than natural gas heating?

Marissa Hummon: So those regions, one thing we've discovered is that the thermal storage properties are more transients in the winter basically because there is a higher temperature differential between inside and outside of the house in the winter. It's more like you know 50 degrees versus 30 degrees or 20 degrees in the summer. And so the heating load shifting is more like two to four hours. But the good thing is is that that corresponds well with the dual peak of heating in the winter season. So generally there's a peak between 7 and 9 AM and again between 4 and 8 PM at night. And so you can actually do what you call you know double load shifting during the day. So you do a little bit in the morning... A little bit of load shifting in the morning and again a little bit of load shifting in the evening. But we have not done any field tests on that so those are all simulation studies.

Chris Nelder: So it's hard to quantify that.

Marissa Hummon: It's hard to quantify down yeah.

Chris Nelder: OK. So you're actually done quite a lot of work I mean as I as I look back over the papers you've contributed to and so on. You've done a lot of work on optimizing DERs for integration into wholesale markets. You've delved into some very arcane issues on how we model that stuff. And I don't claim to understand all your research in this area but I guess I would describe it generally as trying to develop more accurate estimates on the cost of DERs, the production, their potential value to the grid, their ramping characteristics and so on. You know, just this kind of broad effort to really get more granular data on us or be able to model it in a more granular way. Can you explain a bit why so much work has been needed on this kind of modeling and how those findings might be applied?

Marissa Hummon: Sure. So I think you know the guiding principle of my research over the last 10 years has been that optimization works. Everything else is just an approximation to the right answer. So when I say optimization I mean basically linear programs that are used to solve complex problems. It sounds a little ivory tower-ish and I apologize for that. But the field of linear optimization has been prevalent in a wide variety of industries for more than 70 years. You know the electric sector picked up economic dispatch which is a form of linear optimization in the 1950s. And since then I've made a number of big improvements to the use of those models that have had a substantial impact on the total cost of producing power and so a few of those are in about 2003, system operators went from using a strict linear program that would solve basically for the best dispatch point of every generation asset on the system to solving for both whether or not a unit should be on or off based on startup costs as well as the dispatch point. And so when they switch to those mathematics it is called mixed integer linear programming, they suddenly saw that they had a lot more efficient system because they were taking into account those startup costs. Previously they just thought it was good enough to only look at those variable costs like fuel costs. The next major advancement in optimization is in AC optimal power flow. So right now what generally happens when they dispatch a system is they use a linearized version of that power flow equation called the DC power flow and that basically you know it kind of brushes over some of the transient effects of you know different power injection quality points. And so when we master those mathematics of AC optimal power flow we will see another big jump in efficiency. And as I was basically you know learning about all of these big changes in the optimization and working on the issues of distributed energy resources I realized that in order for DERs to be part of the grid they had to basically meet the generation resources where they're at mathematically. And so I basically you know set out to try to represent those system assets in production cost models, in economic dispatch models in order to figure out how they would be dispatched alongside all their generation resources. So while yes it was arcane and it might seem to others as maybe you know do you really need this level of detail in order for this to work? I think that in the long run we'll find that as long as you aren't using optimization to dispatch a set of resources you are not using your system most efficiently.

Chris Nelder: Wow. You just made my head spin like 360. That's incredible stuff. Thank you so much Marissa. It was really fun to have you on the show and there's a lot of really kind of interesting and new ideas to me here so I really appreciate you taking the time to talk.

Marissa Hummon: Well it was my pleasure. And I appreciate all the good conversation and good questions.

Chris Nelder: Sweet. We'll have to have you back on the show in the future to talk about some of the results of the stuff that you've been doing at Tendril. I think that would be fun.

Marissa Hummon: Sounds great.