Skip to content

Why Don’t We Do It With Demand?

(Title inspired by the Beatles. Blog inspired this song sent by an anonymous reader.)

Customers can help avoid blackouts if they are given better information and prices.

Even if you don’t live in California, you probably heard about the rotating power outages we had on August 14 and 15 due to insufficient electricity supplies. While the outages were less than 1000 MW when demand was over 40,000 MW, and lasted for 1-3 hours, any gap in reliable electricity is a problem. But these blackouts were just as important for what they say about California’s electricity planning process as for the direct hardship they imposed. One thing they highlight is how little progress we’ve made towards incorporating demand responsiveness in the transformation of our electricity system away from fossil fuels.

The outages show what can go wrong if we don’t pay sufficient attention to the new constraints from changes in the grid: increasing peak electricity usage due to climate change, more solar and wind generation, closing gas-fired plants and nuclear plants, and, in the future, widespread electrification of vehicles and homes. Some critics are saying the blackouts demonstrate that California can’t pull off a renewables-dominant grid, but that’s the wrong lesson.


Given what a simplistic, rule-of-thumb based approach California has taken to assuring there is sufficient capacity on the grid (known as “resource adequacy”), it’s actually surprising that we haven’t had any system-wide energy shortfalls in the last 19 years. During that time, we’ve ramped up to getting about one-third of the grid’s energy from renewables while also accommodating widespread adoption of customer-side solar and beginning to grow electric vehicle charging. But to keep that momentum going, we need to redesign the way we do resource adequacy planning. We must take a much more granular and empirically-verified approach to planning for supply/demand balance every hour of the year.

There are many ways to make the system balance with high levels of renewables. We could build a lot of storage, including batteries and/or pumped hydro storage. We could keep gas-fired plants online for very occasional use. We could build nuclear plants, like in South Carolina and Georgia. We could do a little of each of these – OK, not the nukes, at least until they come up with something much more cost-effective than the boondoggles in SC and GA.

But the fastest and cheapest contribution to keep supply in sync with demand when the weather is variable and much of generation is intermittent is to reshape demand to more closely track supply. California policymakers talk about demand responsiveness, but efforts so far have been pretty halfhearted.

The most prominent attempt to reshape the state’s demand when there might be a supply shortage is the Flex Alert program, in which the state puts out pleas for voluntary conservation. The state has used this program during the roughest days this month, and it probably helped, though it is difficult to tell intentional conservation from changes in the weather or other factors. The best research suggests that these sorts of voluntary conservation measures can be somewhat effective if used infrequently, but lose their punch with repeated use. I’m not aware of any credible studies on the impact of California’s Flex Alert program. if we are going to continue to rely on it, we need reliable evidence on what it delivers.

Get Pricing Right

Research on demand programs suggests that a better, and more long-lasting, approach is price incentives. Critical peak pricing (CPP) programs give customers lower prices throughout most of the year, but impose a much higher price when supply is tight. Numerous careful studies, covering both residential and commercial customers, have demonstrated that CPP yields substantial demand reductions in response to the high price. And one study by Catherine, Meredith and co-authors demonstrates that making CPP the default gets very high participation and also high satisfaction among customers.

[To respond in advance to those who argue we should pay customers to reduce their demand rather than charging them for peak usage, I addressed that problematic approach in a 2014 blog, which I stand by today. Back in 2009, the Market Surveillance Committee of the California Independent System Operator also evaluated this approach and came to a similar negative view.]

Getting customers to make small demand changes is a cost-effective way to reduce the need for additional grid hardware, whether it is more generation, more storage, or more transmission. California’s demand on the hottest days of the year is about 50% higher than on a typical summer day, and that difference is virtually all cooling. If the outdoor temperature is 95 degrees, it takes about 20% less electricity to keep a building at 78 degrees rather than 74 (details of calculations in this blog are at the end).

And the savings are even larger if you keep the house at 74 until the critical time period begins and then let it gradually drift up to 78. Or, better yet, pre-cool the building to below 74 prior to the critical time and get even more reduction during the peak from this thermal mass form of storage. When you do the math, those sorts of changes in cooling could cut more than 3000 MW, probably much more, when California needs it most.

But to make demand adjustments work, we need to give customers the right signals. It’s time to change the antiquated time periods used for some critical peak pricing, or even for messages about conservation. Many of the programs encourage demand reductions until 6 PM or 7 PM, but that is exactly the wrong time for customers to crank up electricity usage. With so much solar generation now on the grid, the biggest supply crunch comes later in the evening as the sun is setting, typically between 6 PM and 8 PM in the most challenging months, July through September. On most hot days, the best conservation window is more like 5 PM-9 PM than the 2 PM-7 PM that we were hearing from many authorities last week.

And Get Communication Right

Then we have to make clear to customers what actions are important. Air conditioning matters, as does shifting use of electric clothes dryers, dishwashers and ovens, and shift EV charging. Unplugging your cell phone or ipad doesn’t.


Over four hours, changing your A/C setting 4 degrees will probably cut electricity usage by more than 2 kWh. Not charging your phone will save about 0.03 kWh, and unplugging your charged devices will save, well, about zero (Sorry, but it’s true. Test it out with your own watt-meter.) So why are these often put on the same “things you can do” list? (Along with keeping the refrigerator door closed? What household without teenagers needs to be told that? And what household with teenagers would be helped by that suggestion?) 

As for turning off lights, it’s always a fine idea if no one is in the room. But over the four-hour peak period, four strong LED bulbs (surely you don’t light anything with incandescent hogs anymore, right?) use about 0.2 kWh.

We will still need more storage, and some of the gas plants will play a critical role for at least another decade, but reshaping demand is the most agile strategy. That’s particularly important given the uncertainties that climate change is throwing at us. It is time to take demand response as seriously as we take the hardware solutions to grid reliability.

DISCLOSURE: I am a member of the Board of Governors of the California Independent System Operator (CAISO). CAISO dispatches power on the California grid and was the entity that required utilities to institute outages when supply ran short on August 14 and 15. CAISO does not determine or make investments in generation and is not part of the policy process that sets retail rates.

I’m still tweeting energy news stories/research/blogs most days @BorensteinS

Keep up with Energy Institute blogs, research, and events on Twitter @energyathaas



Showing my work

“If the outdoor temperature is 95 degrees, it takes about 20% less electricity to keep a building at 78 degrees rather than 74”. In steady state, the amount of energy required to keep a building X degrees below the default level is linear in X, as explained here, which I confirmed consulting with two engineers. There is some subtlety about the default level, based on sun exposure and thermal mass of the building, but a simple calculation would use the outdoor air temperature. So, energy savings would be 1-(95-78)/(95-74)=0.19.

“When you do the math, those sorts of changes in cooling could cut more than 3000 MW, probably much more, when California needs it most.”  Assume that all of the load between a 32,000 MW weekday and a 48,000 MW week day is cooling and the calculation in the previous paragraph applies. Then, 0.19*(48000-32000)=3040. This does assume that demand response comes from all of that incremental cooling, but it ignores the fact that much of the load at 32,000 MW is also from cooling and  could also respond to peak pricing (as could as many other uses). And, as mentioned in the blog, all of this understates the savings if buildings are kept at 74, or pre-cooled to an even lower temperature, and allowed to drift up to 78 during the beginning of the critical period.

“Over four hours, changing your A/C setting 4 degrees will probably cut electricity usage by more than 2 kWh.” If a customer has central air conditioning, a reasonable estimate is that it pulls 6000 W when operating. I assume that it would be cycling on for half the time during the critical period – likely to be an underestimate during the worst days of the year – which implies 3000 W over four hours or 12 kWh. 0.19*12=2.28 kWh saved.

“Not charging your phone will save about 0.03 kWh, and unplugging your charged devices will save, well, about zero”. My iPhone 6 (yes, I know it’s time to get a new one) uses 7w when charging. So, even if it charged for all four hours of the critical period, it would use 28 Wh or 0.028 kWh. But that overstates the potential savings, because even if my phone is nearly dead, it  reaches full charge in less than four hours. As for leaving charged devices plugged in, I tried this with both my iPad and my laptop computer, as well as my phone. In all cases it was 1-2 Wh per day or less. Now, getting your idle computer, and particularly your screen, to go to sleep when they aren’t in use, that actually amounts to something, about the same as turning the LED lights off in a room.

“But over the four-hour peak period, four strong LED bulbs use about 0.2 kWh.” Replacements for 75 W incandescent bulbs typically use 13w or less.  4bulbs*13w*4hours=204 Wh or 0.204 kWh.


Severin Borenstein View All

Severin Borenstein is Professor of the Graduate School in the Economic Analysis and Policy Group at the Haas School of Business and Faculty Director of the Energy Institute at Haas. He received his A.B. from U.C. Berkeley and Ph.D. in Economics from M.I.T. His research focuses on the economics of renewable energy, economic policies for reducing greenhouse gases, and alternative models of retail electricity pricing. Borenstein is also a research associate of the National Bureau of Economic Research in Cambridge, MA. He served on the Board of Governors of the California Power Exchange from 1997 to 2003. During 1999-2000, he was a member of the California Attorney General's Gasoline Price Task Force. In 2012-13, he served on the Emissions Market Assessment Committee, which advised the California Air Resources Board on the operation of California’s Cap and Trade market for greenhouse gases. In 2014, he was appointed to the California Energy Commission’s Petroleum Market Advisory Committee, which he chaired from 2015 until the Committee was dissolved in 2017. From 2015-2020, he served on the Advisory Council of the Bay Area Air Quality Management District. Since 2019, he has been a member of the Governing Board of the California Independent System Operator.

38 thoughts on “Why Don’t We Do It With Demand? Leave a comment

  1. Thanks, Severin. You’ve convinced me that CPP makes a ton of sense (exempting CARE customers, and perhaps going beyond that by using census data to exempt broader swaths of the low-income population as Scott Burger advocated in his MIT dissertation). But I also want to flag another oft-overlooked part of the solution: a more diverse set of renewables and, in particular, a better wind-solar balance. As several E3 studies have shown (and I expect this to be shown again soon in a forthcoming SB 100 study), relying heavily on solar and storage costs a lot more than dramatically reducing the need for storage through a more balanced renewables portfolio. Then CPP would need to kick in far less frequently.

  2. The theme of the article is solid – it is always less expensive to save energy than generate electricity from any source. California has all the renewable options, so aside from wind and solar photovoltaics, has biomass (biogas from agriculture, food processing, sewage, water treatment plants etc). marine energy, geothermal, and concentrated solar power. And you correctly added storage highlighting batteries and pumped storage, but also should add molten salts, compressed air & liquids, weight, flywheels, and even hydrogen. Already, 43 geothermal plants exist in California, many of which were built in the 70s and 80s and have been generating clean power ever since. Because the state is geologically active, California has great potential for generating power through geothermal plants—according to a 2008 estimate from the U.S. Geological Survey, the state has the potential to produce about 15,000 megawatts of power from geothermal energy (Nat’l Digest 1/23/2020) as an example. But just as sensors, controls, algorithms & aps have enabled hybrid renewable/storage microgrids, the same technologies enable instantaneous load reductions in buildings, specific electric loads, and even EV charging. Such reductions are instantaneous and would have huge implications in easing, and yes, resolving California’s and the other Northwest states summer electricity outages. We need more articles like this with even more specificity. Thanks you for starting this important dialogue.
    Scott Sklar
    Adjunct Professor & Energy Director
    Environment & Energy Management Institute (EEMI)
    and Acting Director, GWU Solar Institute
    The George Washington University (GWU)
    Personal email: Ph 703-522-3049

    • It is not always better to conserve. For example when renewables over produce power and energy it is better to go ahead and consume that energy than to throw it away. This means that new load is needed when solar reaches ever higher output levels. The perfect load for this solar is transportation which already has its own energy storage built into the EVs. So imagine a world where all the EVs are 100% powered off the grid from solar panels that directly plug into those vehicles. You charge at home and you charge at work and you charge in covered parking lots, you charge your cars off direct sunlight from solar panels. I mowed my yard yesterday directly off solar panels to my standard corded lawn mower, no inverter and no batteries needed. This is because these corded 120 vac mowers have series wound motors which run off dc as sell as ac. If you connect solar panels in ah 3s9p format you get about 110 volts open circuit and for panels that produce 240 watts each the mower will draw less than a kw power and run fine even when overcast. I run this off grid. Start thinking about running an entire fleet of electric cars off grid.

  3. Thanks Severin for your relentless banging of this drum! I hereby confess my first failed Ph.D. dissertation topic was on smart metering and time-of-use pricing. I’m not prepared to reveal the year, but needless to say, it was before any customers actually had smart meters. The fundamental issue really hasn’t changed much since then, while we’re still talking about needing better technology. Actually, I think it’s a mistake to get too hung up on the technology of price delivery and response, which is somewhat of a distraction. There are numerous ways of delivering a sequence of prices, which is enough for most customers: text, email, phone calls, radio announcements, even post cards (remember them?), … need I go on? What we’ve ben lacking all along are the obligatory real-time tariffs that would make customers pay attention to those media. Here we are, still chewing on this bone.

  4. Thank you Severin. Excellent article – but that song has been stuck in my head all day. :/

    Based on many years of my own and others’ research with customers, rates and technologies, I’d recommend:
    1. Educate customers about the coming TOU rates and what they can do to minimize their bills, e.g. better insulation+precool, more efficient AC, whole house fan, avoid peak showering with your snazzy new heat pump water heater, etc. Too many to list here. Please DO charge your phone and let your kids watch tv.
    2. Mass-market weatherization blitz by utilities/state, to maximize precooling effects
    3. While your team is there installing insulation, offer low or no cost customer-controlled TOU automation for AC and electric water heaters. Have well-insulated homes precool to 70-72 starting at wake up, and bump up a couple of degrees every few hours. Then set AC to as high as the customer wants when the peak starts at around 3 pm.
    5. TOU rates hit. Customers practice their new skills, watch their new automation automate, feel cool, save money and start to understand and internalize the diurnal cycle of the grid.
    6. Utilities offer CPP to newly aware customers, maybe even RTP to those with proper automation.
    End Goal: Limited load control programs remain in place but are (almost) never used because the price response is sufficient.

  5. At least in California, one of the biggest problems with CPP (and even TOU) pricing is figuring out what the price actually should be. For the IOUs, the rate is roughly half for generation and half for distribution, but few consumers understand that there are two parts and what that distinction means. (And that’s not unexpected because we generally don’t decompose other prices either.) So a CPP rate would affect only half of the electricity rate. But the bigger complication is determining how much the generation rate is and the relationship between the “marginal” price and the average cost that is the basis for setting the utility rate. The “marginal” cost right now is about half of average cost. For PG&E the generation is 11 cents per kWh or about $110 per MWH. State law (AB 57) requires that IOUs be fully compensated for all of their committed generation costs, so any mechanism that does not deliver a full revenue requirement is illegal at the moment. And on the other hand, consumers are strongly resistant to large scale lump sum bill payments that might constitute half of the monthly bill. So this is a political question about what’s feasible given these many constraints and realities. These push California’s rate structure a long ways from an optimal pricing algorithm, so it’s not just a matter of tweaking the system.

  6. At Polaris Energy Services, we just completed a CEC EPIC study showing how dynamic pricing can unlock significant demand flexibility in the agricultural sector AND how the current programs and signals are not driving the desired behaviors. This was proven out last week when we delivered 60 MW of ag load when called upon BUT much of it was during the wrong hours and many more MW under our management were simply not in the market because of bad program design.

  7. Enthusiastically endorsing all of what Sev’s written (and all of what Jim Lazar contributed), plus noting that it won’t work as well as it can without making sure that the price signals that people are offered actually reflect the true marginal cost (i.e., opportunity cost) of wholesale energy. I am awaiting confirmation, but my understanding is that wholesale prices reached only a little above $1,000/MWh at a only few nodes on the system at the time CAISO was implementing controlled, rotating load shedding. The marginal cost of energy during periods of reserve shortages is not, as commonly claimed, the short-run production cost of the last generation unit in the merit order, it is the opportunity cost of using resources to produce more energy (or to use less energy, in the case of demand-side resources) rather than keep those resources in reserve. That opportunity cost is by definition the cost attributable to the increased probability of having to shed load to keep the system up and running, a probability that has obviously reached 1.0 at such time as the system operator must resort to load shedding. Dozens of studies from around the US and around the world of the value consumers attach to avoiding service interruptions, converge on an upper range of about $25,000/MWh. Below that level, however, is a virtually infinite range of values attached to a wide range of electrically-fueled energy services many of which feature at least some degree of controllability and discretion about when they are consumed (or, in the case of stored hot water for instance, when the “fuel” is consumed). Prices of $1,000/MWh would capture the lower end of the range of what customers might happily do to save money while helping to keep their and everyone else’s lights on, but such prices obviously fall well short of signaling all potentially beneficial consumer responses. Once these less-than-transparent real-time prices are filtered through block ToU price mechanisms like CPP or CPR, the strength of the signal, and indeed the amount of compensation available to willing and able consumers, are further watered down. In ERCOT prices can, and on very rare occasions do reach as high as $9,000/MWh, and in part as a result ERCOT has an enviable record of fore-signaling system stress and seeing willing and able market stakeholders throw everything including the kitchen sink at responding. Enviable as well, since again partly as a result of allowing such “shockingly” high scarcity pricing ERCOT has among the lowest wholesale electricity prices in the country, and ERCOT has never had to shed load during the record peak load events the system has experienced in recent years. (Note California load during the recent events was well below the record peaks on the system.) And finally, on the tired and soundly debunked complaint about “equity,” there are a hundred ways to solve that problem to the extent the problem even presents itself in the first place, which as others have noted, is at best debatable.

  8. An unreliable grid is just another reason we should not prohibit natural gas appliances. Last year during the PSPS those of us with gas could cook and have hot water etc. Reliabililty has to come first.

    Distributed storage is one way to improve reliability. If people with EVs or solar/battery combos, can sell their power at CPP prices, it would be an incentive to get such (larger) systems. Such a tide lifts all boats. For the aflluent, it is a fincial incentive and for the poor it reduces the burden by providing more supply when needed.

%d bloggers like this: