He said what?

The consistent and infuriating misunderstanding of a common energy system mechanism

Last Monday the AFR Energy & Climate Summit kicked off with a panel discussion that included the CEO of Boral, one of the biggest manufacturing businesses in Australia. Vik Bansal asserted that not only does Boral shut down their manufacturing plants because of high power prices, but there’s no certainty because the price “…fluctuates every five minutes, with no certainty.”

This was claim was picked up and expanded upon as proof the energy system is in disarray, the “stuttering energy transition”, and of the pressures faced by the largest manufacturers operating in Australia.¹

Except, there's rather a lot more to this story once you dig a little deeper². So let's talk about demand response – what it is, and what it isn't.

Responding Demand What?

The definition of demand response is:

A customer voluntarily reducing their demand (electrical consumption) in response to some external trigger, with associated compensation.

Who the customers are, how much demand is reduced, how that reduction is implemented, what the trigger is and how much the compensation is worth lead us into details.

Now, you’ve probably read variants of the definition above. But I'm firm in my belief that what distinguishes DR is the existence of an external trigger.

The cousins (siblings?) of demand response without an external trigger are demand management and energy efficiency.

Demand management involves customers reducing their demand in line with some kind of economic incentive (price signal). But demand management is self coordinated and is usually fairly consistent – like a customer on a time-of-use tariff who chooses to use less energy during the daily peak window, or the demand management activities associated with most C&I demand-based network tariffs.³

Energy efficiency is terminology usually associated with one-off efficiency improvements which have lasting impacts – installing LED lights throughout a factory will permanently reduce the amount of energy consumed (and save money).

Examples of DR

There are four fundamental types of demand response:⁴

1. Economic DR

2. Capacity DR

3. Network DR

4. Interruptible load

Let's talk about demand response, with examples from the NEM.

Economic DR

Economic DR can be explained very simply — demand response for economic reasons. But remember that there needs to be an external trigger for DR, so economic DR is triggered by a strong economic incentive not to use load (not just the daily metronome of time of use tariffs).

Let's take the example of economic DR in the NEM.

The NEM is an energy-only spot market with a floating 5 minute price. Of the half-dozen energy-only markets around the world the NEM has one of the highest caps at ~$16,000/MWh.⁵ We could spend several paragraphs talking about this in more detail, but the one liner is — most of the time the price is pretty 'normal' and every now and then it spikes several orders of magnitude higher (volatility).⁶

If there are only a handful of volatility events in a given year (hours, or perhaps tens of hours over the course of a year), then you could reasonably choose to pay this floating spot price, knowing that there will be some painful periods relative to the average price.⁷ Better yet, if you can do the demand response thing, by reducing your demand when the price exceeds some threshold (the trigger) than you’ll end up paying lower than the system mean — you'll be paying the average of only the non-volatile periods.

Why would someone choose to take such a seemingly risky gamble? Because fixed prices offered by retailers inherently have some measure of this volatility built into them — you’re paying for the retailer’s estimate of the value of that risk (and your ability, or inability to reduce that risk burden for the retailer).

It’s akin to paying for the gold standard in health insurance. If anything ever goes wrong you're safe in the knowledge you’ve got great coverage. But it’s not an unreasonable strategy to not spend that money on health insurance and take the risk that nothing bad happens (and perhaps you could take the money not spent on health insurance and keep your own emergency reserve?).

There are a few variants on this strategy, but the outcome is fundamentally the same — when [spot price] line go up, load go down.

The Wholesale Demand Response Mechanism is a formalised mechanism which allows large users to participate in economic DR, without having to take a fully spot exposed retail contract. Even more significantly, the mechanism allows for third party companies — not the customers or the retailers, but aggregators — to be the organiser of that DR.⁸

Outside of the WDRM, or spot exposure retail contracts, retailers are free to offer their own variants of economic DR with their customers. Customers pay some kind of fixed price, but are either rewarded with discounted rates or a revenue stream for participating in demand response when the retailer calls it.

Capacity DR

What if a pricing signal isn’t the primary driver of demand response? What if instead of paying generation to turn on to meet high demand (system peaks), loads are paid to reduce that peak demand?

That’s the fundamental premise of peak or capacity DR — there’s an argument that it might be more economically efficient to reduce the system peak than it is to build the generation in the first place (and connect it to the grid and maintain and operate it, etc.).

The devil is of course in the detail here; there’s a reason system peaks occur in the first place, so not all loads will be suitable or able to provide this kind of DR.

The other subtlety is that these loads might not normally participate in the market with economic DR, but will do so given a sufficient pricing signal (i.e. more money then they could earn via economic DR). So capacity DR schemes have the potential to bring DR capable loads out of the woodwork, especially if the commercials are good and the trigger is not pulled too often.

This is the basis for the Reliability and Emergency Reserve Trader (RERT) mechanism. Loads (and also generation) which are not typically active in the market can be paid to provide DR as a measure of last resort. AEMO keeps a RERT portfolio on standby for use on the very hottest days in summer when there may not be enough generation to meet the system demand.

There’s a genius here too — the loads (factories, warehouses, manufacturing, industrial facilities) and the power lines connecting them already exist. The payments are to pause production for a period of time and be compensated for it.

Network DR

Network DR follows a somewhat similar concept to capacity DR, however instead of system peaks, networks seek to manage peaks within their own network patches. These might be coincident with system peaks, but equally might not, especially if only specific areas of the network are affected.

Network DR is usually driven by the desire to defer capital expenditure on upgraded network equipment (augmentation). If DR could reduce the peak load seen on a feeder or zone substation, then investment in upgrading the equipment could be pushed back, maybe.

In practice, this type of DR is one of the hardest types of DR to get right from a program design perspective.

Consider that most networks operate as regulated utilities — they are incentivised to increase their regulated asset base, upon which they receive a guaranteed rate of return. In Australia there are mechanisms to encourage the deferment of capital expenditure where possible via the euphemistic ‘non-network solutions'; but it’s hard to make the economics work if you’re probably just going to spend the money anyway in a couple of years.

One of the more notable network DR programs in the NEM is the AusNet Critical Peak Demand program in eastern Victoria, which defines five days during summer and uses network tariffs for large businesses to heavily incentivise demand response during these windows. It heavily blurs the line between demand management and DR, but we’re calling it DR.

Interruptible Load

I’ve saved my favourite DR for last — interruptible load. These are loads which can incredibly rapidly (in a matter of seconds) and without warning be reduced in order to respond to changes in the system frequency.⁹ The rapid reduction of load (equivalent to injecting more power) on the system helps to stabilise the frequency and prevent potential system collapse.

In the NEM there was a rule change in 2017 enabling large loads (> 10MW) or aggregated loads above this threshold to participate in the contingency frequency control (FCAS) markets. Loads now consistently make up more than a quarter of all provided contingency raise FCAS.

The kinds of loads that can provide these responses are somewhat niche — can you handle having the power to your factory or industrial process switched off without any warning (and without any kind of predictability)?

The silver lining is that typically this form of DR is the most valuable, certainly the case in the NEM.

This is a baseline joke. But I’ve run out of time to actually talk about them. So just enjoy this out of context meme.

I want to wrap up with a story about the most interesting DR program I ever heard about. The town of Danvers, Massachusetts (part of the ISO New England RTO) implemented a residential DR pilot program in an extremely low cost manner. The organisers arranged an emailing list of residents using Mailchimp. When there was a DR event predicted they emailed customers several days in advance advising them of the event and the actions they could take to reduce power consumption. After the event they sent around a survey (Survey Monkey, of course) asking what actions customers had taken. If customers said they had done two or more things, they mailed out $10 Amazon gift cards.

No metering, no control rooms, no IT systems integration. Very low overheads. Now this program is obviously prone to gaming from dishonest (but economically rational) individuals, and is not especially sound from system security perspective. But it was an interesting experiment in how behavioural DR could be managed.

What is DR not

Let's go back to the definition of DR:

A customer voluntarily reducing their demand (electrical consumption) in response to some external trigger, with associated compensation.

Demand response is always a compensated and voluntary thing; every demand response program I've ever seen the program for have optional response clauses in them. Optional responses might attract penalties or loss of revenue, but participation in demand response is a contractual arrangement with the ability for customers to opt out. Customers sign a contract with agreed rules of engagement and the promise of some kind of compensation.

Demand response is not load shedding. Load shedding is an involuntary and emergency mechanism to deal with system security issues. No one wants to see load shedding. Not customers, not the system operator, not the regulator.¹⁰

Load shedding is a safety mechanism built into electricity systems. The fundamental idea is that if the energy system can't handle the load (demand) on the system, it's better to kick a smaller handful of customers off, then risk a domino effect which would catastrophically black out much larger swathes of customers (or entire grids).

Currently Speaking does not condone violence. But don’t be confusing load shedding and DR, Batman is a masked vigilante.

It’s all worth noting that I think DR is a valuable and economically efficient mechanism in any electricity system. But that doesn't mean it's not without its problems or challenges.¹¹

Some of these problems are down to the design of the specific program and incentives to customers; others relate to how the customers themselves engage with the program and organise internally.

But whatever its flaws DR is certainly no longer a novelty, and is an increasingly valid way to deal with the modern grid.

Restoring Load

Returning to Bansal's comments, the description he gave of Boral’s operations was consistent with a spot price exposure contracting strategy, a not uncommon strategy among big sophisticated energy users with the resources to manage it and loads flexible enough to handle it.¹²

And although he doesn’t like the uncertainty associated with 5-min spot exposure, he baulks at the price of fixed contracts and doesn't want to be locked in to long term deals (which would definitely provide certainty).¹³ Honestly, this is having your cake and eating it too. The entire retail market is literally based on short to medium term fixed price contracts which provide price certainty, he just doesn't like the sticker price.

As to the headline and copy in the opening paragraph, spot price responsive demand response has always been a feature of the NEM. In fact there's a strong argument to be made that after twenty years we still don't have the kind of price responsive flexible load in the system as envisioned by the original designers.

So claiming it as an outcome of the 'stuttering transition' is cheap, lazy and attempting to jump on the bandwagon of the current [legitimate] energy system concerns. Do better AFR.

---

Things Happen

• The shuttered White Bay Power Station in inner Sydney will play host to an arts festival next year. Next stop, theme park 🤞

• The ACCC has officially approved the sale of Origin to Brookfield, but now the shareholders are grumpy the deal isn’t good enough. We wait, we watch.

• ABC Four Corners will air a piece on Snowy Hydro 2.0 on Monday 23 Oct. Finally the nation can hear Florence the TBM’s side of the story.

1. The AFR piece was honestly about as well researched as a PR release, which is disappointing given the AFR generally has some of the best energy coverage of the mainstream media. We’re not even going to talk about the dogshit written by this perennial climate cretin.

2. Like, scratch off the top layer of dead skin.

3. There are, of course, grey areas. But I still think the nuance is worth maintaining.

4. As best I know these are the same four fundamental categories in all energy markets, but I’m willing to be corrected here.

5. There's a floor of -$1,000/MWh too. No, that’s not a new thing.

6. In fancy statistical terms this is called a stochastic (random) mean-reverting system — the volatility is random, and when the volatility is over the system price returns to normal.

7. Also despite the system being considered stochastic, there is a reasonable level of predictability if you know what to look for.

8. This might seem insignificant, but in a market dominated by three vertically integrated gentailers, it’s not. When it was introduced it finally signalled a shift from the regulator accepting the competitive role these companies could bring to the market.

9. Technically you can provide an increase in load for managing system over-frequency events too, but let’s leave that discussion for another day.

10. Maybe conservative columnists, because they’ll finally get to say “I tOlD yOu So!”. Presumably it hurts endlessly predicting blackouts for them never to arrive.

11. No, one of those problems is not that asking customers to reduce electricity usage is fundamentally 'bad'.

12. Boral are one of the few companies to manage their own interruptible load FCAS program outside of an aggregator; they’re not amateurs.

13. To be fair this is straight out of the aluminium smelter playbook. Alcoa has been bitching about power prices for as long as I’ve been alive, doubly so when it’s recontracting season!