“Dispatchable” As a Product Characteristic, and How to Deal With It in Markets

Lynne Kiesling

Derek’s comment about wind power on my previous post mentions that wind power is not “dispatchable”. This characteristic refers to the extent to which the resource is available to the system operator, who is responsible for the real-time physical balancing of the network, when striving to achieve that balance. Dispatchability has been a serious issue in regulatory policy discussions, and I think it will be even more important in the future.

Here’s the issue: control room operators are responsible for making sure that the power flows on the grid keep voltage and frequency levels within particular ranges to maintain power flow on the grid. This task is complicated, and the physics behind the task is even more complicated. So if you are a control room operator, you want to have a sizeable portfolio of resources at hand that you can turn on and off, and that you can know will respond. If you are a control room operator, you want certainty, including certainty that when you tell someone to power up or power down, they are actually going to do it in the timeframe you requested.

Control room operators (and, by extension, system operators) have issues with both supply-side and demand-side dispatchability of resources. Derek’s comment points to the supply-side issue with wind; the physical challenge of incorporating wind into the grid is that you have to produce power when the wind is blowing, and you can’t control that. It’s not like powering a gas turbine up or down, which is much more amenable to the centralized control paradigm in which the control room operators have traditionally functioned in the regulated environment.

Dispatchability on the demand side is a topic of particular, and timely, interest to me. The control room operator is not a big fan of demand response of the market-based, commercial type that I advocate, because it increases his uncertainty. If all customers can respond to prices, the control room operator loses the ability to control demand from his centralized vantage point. Thus from the control room operator’s perspective, decentralized demand response enabled by market contracts removes a tool from his toolkit. How can he control when “load” is going to come on and off? At least with regulated, averaged, fixed retail prices, “load” is more predictable, and follows historical load duration curves pretty well. Once individuals have the freedom to choose how, and how much, to respond to prices, those load duration curves may no longer hold.

So there’s the policy challenge, and the persuasion challenge for those of us who argue that market processes enable intertemporal, dynamically efficient resource allocation and production. We have to build the case for how market processes and customer choice provide decentralized control distributed throughout the grid. Furthermore, we have to be able to substantiate the claim that such distributed control will actually make the control room operator’s job easier, not harder.

Consider two different factors that contribute to that claim. First, put on your new institutional/transaction cost economics hat. If we think at the transactional level, about a wholesale power transaction between a generator and a load-serving entity (as an aggregator/proxy for their end-use customers), one way to mitigate some of the uncertainty facing the system operator is contractually to specify dispatchability as one dimension of the transaction, with the system operator having full information about the planned power flows and the characteristics of the resource. Some resources are more dispatchable than others, and if that dispatchability has value, then a dispatchable resource should command a higher price in the marketplace.

Incorporating that insight into contracts should not be prohibitive. For example, in that transaction you have to pay for the transportation of the power; i.e., you pay for transmission. The system operator sits at the middle of that portion of the supply chain. So the contract for dispatchable power could involve a lower transmission charge than the contract for non-dispatchable power, reflecting the value differences between the dispatchable and non-dispatchable resources. The challenge is to think about how to create an institutional environment in which such value-reflecting transactions and contracts can occur.

The second factor that affects my claim of distributed control is technological change. The centralized control room paradigm is premised on an analog world. Now we have a plethora of digital technologies, including those that enable distributed remote sensing and monitoring, that make distributed, decentralized control possible. The options include automated sensing and response; both demand and supply resources can program sensors to take an action in response to either a voltage change, a frequency change, or a price change. In aggregate, those distributed automated responses may provide emergent order, in the form of voltage, frequency, and price stability. In that sense distributed control can make the control room operator’s job easier, by requiring him to focus only on truly real-time unanticipated events, which may become even less frequent as the distributed technologies and price structures and contracts reduce the probability of cascading failures.

+

10 thoughts on ““Dispatchable” As a Product Characteristic, and How to Deal With It in Markets”

  1. My initial thoughts were along the lines of your second factor. I don’t think that consumers who can respond to real time prices will prove to be less predictable than consumers facing fixed prices.

    If anything, such consumers should be more predictable: tending to reduce consumption in response to higher prices, tending to increase consumption in response to lower prices. It may be that traditional load forecast don’t consider price effects — just looking at weather, day of week, time of day, and other factors — but that lack should be easy to fix with a little study.

    What’s more, price-responsive consumers should tend to act in ways that make the system operator’s job easier. Not just in the emergent order possibility you identify, but even at a more basic level. Often prices will be high during times of system stress (and system operator duress), and price-responsive consumers will tend to back down consumption and ease the system operator’s job.

    Of course, sometimes system problems happen off-peak and often problems manifest themselves faster than can be reflected through markets and prices. But with the automated technologies you mention, the demand side of the system can stand equally ready as the supply side to respond to the system operator’s emergency actions, and such systems should be equally able to get paid for the value provided to the system.

  2. My initial thoughts were along the lines of your second factor. I don’t think that consumers who can respond to real time prices will prove to be less predictable than consumers facing fixed prices.

    If anything, such consumers should be more predictable: tending to reduce consumption in response to higher prices, tending to increase consumption in response to lower prices. It may be that traditional load forecast don’t consider price effects — just looking at weather, day of week, time of day, and other factors — but that lack should be easy to fix with a little study.

    What’s more, price-responsive consumers should tend to act in ways that make the system operator’s job easier. Not just in the emergent order possibility you identify, but even at a more basic level. Often prices will be high during times of system stress (and system operator duress), and price-responsive consumers will tend to back down consumption and ease the system operator’s job.

    Of course, sometimes system problems happen off-peak and often problems manifest themselves faster than can be reflected through markets and prices. But with the automated technologies you mention, the demand side of the system can stand equally ready as the supply side to respond to the system operator’s emergency actions, and such systems should be equally able to get paid for the value provided to the system.

  3. I agree with Mike on predictability. Price response moderates demand in a beneficial way, and probably reduces its volatility relative to no price response. But we shouldn’t fool ourselves about predictability. Demand is not really very predictable now, but the methods that are used for forecasting use historical relationships between weather and demand, for instance. What price response does is reduce the sensitivity of demand to weather. That alone may improve the predictability, but in the end all you do is change the historical relationships in your forecasting models. As Mike says, it’s an easy fix. But there’s another idea we should mention. Operators don’t necessarily control demand very much, or perhaps it’s better to say that they don’t control much demand. On our large system there’s very little direct load control, and a limited amount of industrial interruptible. Interruptible is generally treated as a last resort, although many companies now consider the marginal cost of power when deciding whether to interrupt or not. In any case, those details are spelled out in their contracts. Some companies also have programs where price-sensitive customers are exposed to market prices for a limited number of days per year, at the choice of the system operators. Finally, some companies have had real-time industrial prices for many years, yielding much data for analysis.

    As for the operators, they like having interruptible available, especially if it is interruptible on very short notice. That makes it useful in response to large supply or transmission contingencies, and it stands in place of operating reserves, reducing operating costs even when it is not being interrupted. In a sort of paradox, real-time prices tend to eliminate short-notice interruptible at the times it would otherwise be most valuable. That is, only very flexible customers can offer short-notice (5-minute) interruptible, but if they are both flexible and price sensitive then their demand is reduced at peak times when operating reserves are scarce. There’s a happy medium in there somewhere.

    I worry less about the operators and real-time conditions than I do about day-ahead decisions made by traders, day-ahead portfolio planners/optimizers, and large price-sensitive consumers. Price sensitivity of large customers isn’t necessarily the result of decisions made in the heat of the moment. They are made in a more rational planning timeframe of at least a day ahead. What I would like to see is day ahead transactions where consumers decide on the next day’s production and power purchase based on a day-ahead price, and then adjust during the operating day according to real-time prices. The day-ahead timeframe is much more realistic for price response, and without a real day-ahead transaction we’d probably find the consumers basing day-ahead decisions off of uncertain day-ahead predictions rather than in real time. On the other side of the ball, the traders are having to take day-ahead positions in response to forecasted conditions and prices. Some dickering between price sensitive consumers and day-ahead traders could be beneficial, and even some integrated utilities have programs that offer such an arrangement.

  4. I agree with Mike on predictability. Price response moderates demand in a beneficial way, and probably reduces its volatility relative to no price response. But we shouldn’t fool ourselves about predictability. Demand is not really very predictable now, but the methods that are used for forecasting use historical relationships between weather and demand, for instance. What price response does is reduce the sensitivity of demand to weather. That alone may improve the predictability, but in the end all you do is change the historical relationships in your forecasting models. As Mike says, it’s an easy fix. But there’s another idea we should mention. Operators don’t necessarily control demand very much, or perhaps it’s better to say that they don’t control much demand. On our large system there’s very little direct load control, and a limited amount of industrial interruptible. Interruptible is generally treated as a last resort, although many companies now consider the marginal cost of power when deciding whether to interrupt or not. In any case, those details are spelled out in their contracts. Some companies also have programs where price-sensitive customers are exposed to market prices for a limited number of days per year, at the choice of the system operators. Finally, some companies have had real-time industrial prices for many years, yielding much data for analysis.

    As for the operators, they like having interruptible available, especially if it is interruptible on very short notice. That makes it useful in response to large supply or transmission contingencies, and it stands in place of operating reserves, reducing operating costs even when it is not being interrupted. In a sort of paradox, real-time prices tend to eliminate short-notice interruptible at the times it would otherwise be most valuable. That is, only very flexible customers can offer short-notice (5-minute) interruptible, but if they are both flexible and price sensitive then their demand is reduced at peak times when operating reserves are scarce. There’s a happy medium in there somewhere.

    I worry less about the operators and real-time conditions than I do about day-ahead decisions made by traders, day-ahead portfolio planners/optimizers, and large price-sensitive consumers. Price sensitivity of large customers isn’t necessarily the result of decisions made in the heat of the moment. They are made in a more rational planning timeframe of at least a day ahead. What I would like to see is day ahead transactions where consumers decide on the next day’s production and power purchase based on a day-ahead price, and then adjust during the operating day according to real-time prices. The day-ahead timeframe is much more realistic for price response, and without a real day-ahead transaction we’d probably find the consumers basing day-ahead decisions off of uncertain day-ahead predictions rather than in real time. On the other side of the ball, the traders are having to take day-ahead positions in response to forecasted conditions and prices. Some dickering between price sensitive consumers and day-ahead traders could be beneficial, and even some integrated utilities have programs that offer such an arrangement.

  5. Wouldn’t some of the problems with dispatchability go away if the wind resource was pared with a coal or natural gas unit so that the fossil unit would ramp up or down in response to the intermittancy of the wind resource? This could eliminate concerns of imbalance penalties, etc. and result in a smoother and more predictable flow of power from the wind/coal provider.

  6. My thoughts are that dispatchability is indeed a problem that is screaming out for a storage solution, rather than better “dispatchability”. For example, although the idea of a hydrogen economy is currently more farfetched than supporters would like to admit, the concordance between wind-power generation and hydrogen-creation and storage/transportation seems straightforward. If they can create hydrogen with “extra” energy they can transport that much more easily. Or store it and release the energy when wind isn’t there.

    “Easily” is of course in the eye of the beholder. I’m not an engineer, so perhaps my suggestion is farfetched, but until I hear a good explanation of why, it sure seems like a logical association. If the supply is not “dispatchable” you need to pair it with a storage mechanism. Just a thought.

  7. My thoughts are that dispatchability is indeed a problem that is screaming out for a storage solution, rather than better “dispatchability”. For example, although the idea of a hydrogen economy is currently more farfetched than supporters would like to admit, the concordance between wind-power generation and hydrogen-creation and storage/transportation seems straightforward. If they can create hydrogen with “extra” energy they can transport that much more easily. Or store it and release the energy when wind isn’t there.

    “Easily” is of course in the eye of the beholder. I’m not an engineer, so perhaps my suggestion is farfetched, but until I hear a good explanation of why, it sure seems like a logical association. If the supply is not “dispatchable” you need to pair it with a storage mechanism. Just a thought.

  8. Mark,
    Most coal plants aren’t designed to ramp up and down a lot, but you’re exactly right that gas (and hydro) are well suited for this kind duty. But having to have those resources as a back up to wind is a significant cost.

    So let’s say a utility builds a 100MW-wind farm. The actual expected power out of the thing might be closer to 40MW, let’s say (taking into account the variable nature of wind). But then the utility has to put some existing generation in reserve to backup the new wind plant. So in the end, what looks like a 100MW addition to your portfolio turns out to be something more like 20MW (these numbers are only for example’s sake!)

  9. My air conditioner has a little radio controlled box attached to the power line to the compressor. The box enables the utility to turn off the AC for up to 3 hours per day, during peak periods. For my trouble, I get a $10/month discount on my electric bill.

    What I’ve done here is sell demand dispatchability to the utility. I believe they have similar, if much larger, contracts with larger commercial and industrial users.

  10. My air conditioner has a little radio controlled box attached to the power line to the compressor. The box enables the utility to turn off the AC for up to 3 hours per day, during peak periods. For my trouble, I get a $10/month discount on my electric bill.

    What I’ve done here is sell demand dispatchability to the utility. I believe they have similar, if much larger, contracts with larger commercial and industrial users.

Comments are closed.