Skip to content

The Duck has Landed

May has arrived and days are getting longer and warmer. This is good news for baseball fans, barbecue enthusiasts, and grid operators concerned about integrating unprecedented levels of solar energy onto the California grid.


Source: Solar panels at Busch Baseball Stadium

Plugging lots of solar into the power system creates challenges, particularly on days when electricity demand is relatively low and renewable generation is high. Here in California, this happens in March and April when solar intensity is up (relative to the winter months), but air conditioning demand has yet to kick in.

Back in 2013, some California energy analysts with an eye for aesthetics were looking at how projected increases in renewable energy generation might affect power system operations. They plotted actual and projected hourly net load profiles (i.e. electricity demand minus renewable generation) over the years 2012 to 2020, focusing on late March when integration concerns loom large. The result was remarkably duck-like.


The California ISO “duck chart” made a big splash for a number of reasons. For one, a graph that looks like a duck makes an otherwise dry, technical issue more fun to talk about.  Conversations about renewable integration become more engaging when sprinkled with fowl word plays.

Perhaps more importantly, the graph highlights two related integration challenges. First, the long duck neck represents the steep evening ramp when the sun sets just as Californians are coming home and turning on their lights and appliances. Accommodating this ramp requires maintaining a fleet of relatively expensive generation resources with high levels of flexibility. Second, the duck’s growing belly highlights the near-term potential for “over-generation”. As solar penetration increases, net load starts to bump up against the minimum generation levels of other grid-connected generators, such as the state’s remaining nuclear power plant. At some point, system operators have to start curtailing solar to balance the grid.

How’s the duck shaping up?

The CAISO duck chart predicts that we should see increasingly duck-like net load profiles in March and April. So I’ve been keeping an eye on the great data that CAISO makes readily accessible. This year, the duck showed up. The graph below plots average net load profiles for late March/early April since 2013 (I averaged across seven days around March 31 to smooth out the variation that comes with random weather, week days versus weekends, etc.).

duck_graph.fwNote: All data taken from CAISO website. Graph summarizes hourly data, March28-April 3, 2013-2016.

In the 2016 duck season, we saw mid-day net loads at or around predicted levels. Increased solar penetration on both sides of the meter (utility scale and distributed)  has been driving net loads down when the sun is up. Fortunately,  the ramp from 5 – 8 pm has not been quite as steep as projected because electricity demand in the evening hours has  been lower than projected. Perhaps this is due to unanticipated demand-side energy efficiency improvements. I could not easily find hourly curtailment data. The data I could find on plant outages indicate that March 2016 saw the highest forced solar plants outages on record, but these outages could  be due to factors other than curtailment.

My after-the-fact duck chart suggests that renewables integration challenges are showing up more or less on schedule (although ramping requirements are somewhat less than projected). So far, these challenges are quite manageable without major changes to grid operations. But the duck of the future – especially given California’s new target of 50% renewables by 2030 –  will present a more formidable challenge.

Renewables integration strengthens the case for regional coordination

California is not alone in creating and confronting unprecedented renewable integration complications. Take Hawaii, for example, where a 100% renewables target makes California’s 50% look timid. Our colleagues at University of Hawai’i, Michael Roberts and Mathias Fripp, have been thinking hard about how Hawaii can pull this off at least cost. The charts below illustrate a hypothetical 100% day in Oahu in April (no more duck when all load is served by renewable energy!):


Source: Fripp (2016)

The broken line in the right graph represents the “traditional”, business as usual demand profile. To hit the 100% target, wind and solar generation increases to nearly double current levels of the traditional peak.  Differences between the timing of renewable energy production and traditional demand are reconciled primarily by EV charging and other demand-side response  programs (although batteries and pumped storage also play a role).

When you’re an island in the middle of the ocean, you’re pretty much on your own when it comes to tackling these grid integration challenges. Thus, Hawaii is preparing to demonstrate how significant renewable energy integration can be achieved with demand response, grid management, and storage. In contrast, California has more options to leverage.

Although California fancies itself a different world, it is physically connected to (but not perfectly integrated with)  a larger western power system.   From an economic perspective, expansion of the energy imbalance market and improved coordination of the western grid looks like an obvious and important piece of California’s renewable integration puzzle.  A regionally coordinated western grid would integrate mandated renewables across a larger area, thus reducing the likelihood of over-generation. Coordination across balancing areas should also provide increased flexibility.

In the past, economists have documented the efficiency gains of improved regional coordination and bemoaned the inefficiencies of the balkanization that persists.  Looming renewable integration challenges could provide the needed additional impetus for grid integration.  To be sure, there are some important details that need to be better understood. But if done right, a fully coordinated regional grid could help clip the duck’s wings.


75 thoughts on “The Duck has Landed Leave a comment

  1. The original article talks about the “formidable challenge” of dealing with a 50% renewable system in 2030. To paraphrase William Gibson, the future has already arrived, it’s just not evenly distributed. As I write this (5/15/16, just before 2 pm), the CAISO has renewable generation (not counting large hydro) of 12,869 Mw and load of 22,871 Mw, meaning it is 56.3% renewables. And it has been over 50% for 4 hours today so far (and was over 50% for 5 hours yesterday).

    So while the goal is to average 50% renewables in 2030, meaning we can expect 50% renewables in roughly 4400 hours per year by then, we are already experiencing 50% renewables today, albeit in many fewer hours.

  2. What is clear from Gene Preston’s post is that 330 hours of BATTERY storage will not be a cost-effective option with $400/kWh storage.

    But nuclear is not a solution, because it is also too inflexible — units cannot ramp more than about 2% per hour, so you can’t follow load with them either. And you need to find a market for power in the middle of the night. France exports at night, imports by day.

    Fortunately, there is MUCH lower cost storage available, in the form of thermal storage in water heaters and air conditioners.

    Controlling water heaters is an old science, originally done to move the load off-peak to adapt to inflexible nuclear (France) and coal (US Coops) generation that produce half of the power when you don’t want it. Now adapted to control actively, to provide both diurnal storage and ancillary services, the costs are often negative (i.e., the ancillary services revenue pays 100%+ of the incremental cost for control). Because electric resistance water heaters can be controlled in intervals as short as 4 seconds (no moving parts), this is an ideal resource to smooth out wind and solar variations. PJM and Hawaiian Electric are proving they can follow a frequency regulation signal VERY closely. PJM is contracting for 30,000 water heaters for ancillary services. There are 45 million electric water heaters in the US (about 2 million in California).

    Ice storage air conditioners are basically your home fridge’s ice machine, plus a fan. OK, it’s a little more complicated, but not rocket science. Produced by Ice Energy, CALMAC, and others, these provide low-cost storage, typically at a net savings in kWh (night-time operation of the chillers is more efficient due to lower outside temperatures and better heat rejection. Air conditioning is about half of the US peak demand, so changing this from an as-used load to a scheduled and controlled load is a huge opportunity.

    For other ideas on how to adapt to a bulge of generation at mid-day, see Teaching the Duck to Fly, Second Edition at

    • Jim your nuclear observation is incomplete. There are three kinds of nuclear power in my model. The first is the base load nuclear you describe. The second and third kinds of nuclear are the Per Peterson type of nuclear that are dispatchable, see which has a central high temperature gas reactor feeding 12 mostly conventional design turbines of 240 MW maximum capacity each but running at 100 MW in nuclear only mode. The turbines can be individually dispatched in 100 MW increments nuclear powered for load following. More importantly, these turbines have a gas fired boost of an additional 140 MW for covering spinning reserve requirements. Because the turbines are already hot and running there is no delay and these turbines will be extremely fast acting to cover loss of other generation or even to follow the variations load that renewables create. This means that all the spinning reserved is covered by nuclear generation and gas fired boost. You don’t even have to keep gas generators in standby to cover spinning reserve. This also replaces the need for batteries since the energy is stored in the form of nuclear fuel. Its a wonderful concept that I have programmed into my nuclear generation models. It works great in the models.

      • Gene, you’re proposing to take a technology that is already expensive and making it even more expensive by taking away it’s one advantage of being steady baseload.

      • Didn’t the UK try HTGRs and abandon them for LWRs?

        And would these nuclear units vent the hot gas to the atmosphere? That seems inefficient. And the turbines would be 140% oversized most of the time.

        Also, I think you are conflating storage with unused capacity. Your hybrid nukes wouldn’t store energy, they would just be dispatched. The energy is not “stored in the form of nuclear fuel” any more than energy is stored in coal piles. When renewables are generating more energy than necessary, you can’t store the excess in the nuclear, coal, or gas units. You can do that in energy-limited hydro, pumped hydro, batteries, and other storage technologies.

        • We are talking about a new nuclear plant design. I have modeled it in detail. Here is the specific simulation: There is no venting of hot gas moreso than a combined cycle gas plant. Scroll do the second page from the bottom of that file and see that the amount of gas boost is a very small amount of energy. Its the smoothest running plan. The others have to have huge amounts of storage to handle the enormous swings in renewable power.

          • Gene, You didn’t model the plant design. You ran some sort of dispatch model. Using 1960s graphics. It was like the good old days.

            You are not doing a very good job of marketing Peterson’s HTGR “design,” since you haven’t explained where the exhaust gases go, whether this is supposed to be a combined-cycle or simple-cycle design, how much efficiency would be lost in running a 250 MW turbine and generator at 100 MW 90% of the time, or even how the new HTGRs would differ from the old, failed HTGRs.

            Lots of breakthrough nuclear designs have come and gone without ever being built. I’ll pay attention to this design when a utility orders the first one.

          • I didn’t even run a dispatch model. I simply tested to see if there was sufficient capacity every hour to meet the demand. I’m getting ready to update the IEEE RTS with a wind model and will present my suggestions to the LOLEWG this summer in Boston. I have verified Roy Billington’s 1986 LOLE ‘exact’ calculations of the LOLE to six decimal places. However my model is more advanced this his since I can model loads up to a million MW peak demand and hundreds of thousands of generators, all calculated with an exact process. The LOLE test I am doing simply asks the question, have we built enough capacity to keep the lights on? Look under RTS on my web page for more info including a free program you can use to run your own LOLE calculations on any large system. see The wind I just added to datain.txt is purely a synthetic model. Compare the erratic nature of wind with the real wind erratic profile in Texas on the spreadsheet Note all the gaps in the wind output. This is a real problem with wind. Furthermore wind is now showing signs of deterioration in the elements. See these references and This is not good news for planners.

          • Gene, Thank you for the clarification that you did not model hypothetical new plants, but just assumed that they would be highly reliable. You found that reliable generators would provide reliable energy supply.

            And thanks for the links to studies of wind ageing. We’ll see how fast the technology improves. And whether anyone develops a nuclear design that can actually be built economically.

    • Using water heaters to smooth the variability and uncertainty of renewable production might well be useful, but it doesn’t solve the bigger problem, which is how to power homes and businesses when the sun doesn’t shine and/or the wind doesn’t blow. For that you need either a standby resource (presumably gas-fired) or less renewable production coupled with some other carbon-free resource (presumably nuclear) or lots of storage as Gene points out, because you can’t power a light bulb with hot water.

      Solar and wind are cheap on a busbar cost basis. They’re not so cheap when the cost of dealing with variability, uncertainty and periods of zero production are factored in.

      California has plans for all new residential and commercial buildings to be zero-net energy structures. It sounds nice but as you start thinking about how it would be accomplished, you realize there are a number of practical and institutional problems to be overcome and many of them are not trivial. For example, do we force homeowners to either cut down or refrain from planting trees around their home in order to maximize solar production? How do you deal with high density housing? If I need five or so Powerwalls just to get through a period of reduced solar production due to weather, where do I put 1,000 pounds of batteries?

      • Jack, several answers to your questions:
        – We still have much “storage” in the current grid that won’t go away for a while–coal-fired generation in the rest of the WECC that will be forced to ramp when faced with zero-priced.
        – And future storage need not be single-purpose dedicated use. EVs are likely to be a very cheap source of storage. And we’re more likely to get a breakthrough in hydrolysis than fusion.
        – Community solar gardens solve the problems of shaded roofs and split incentives with tenants. These are easily solved institutional/regulatory problems so long as the entrenched interests cooperate.

      • The more likely case is single-purpose storage. Storage built for frequency regulation isn’t suitable for ramping, storage built for ramping isn’t suitable for “arbitrage”, and I’m pretty sure using EV batteries for grid services is not going to be as easy or as acceptable as proponents would like to think because it will be very difficult to have that storage serve two masters.

        • Jack, what’s to say that auto makers don’t offer different types of storage capabilities in their cars? When we get to very high EV penetrations, we will have way more storage than we will ever use, and that storage will be essentially free because it’s primary purpose will be for transportation. It’s like how land lines became obsolete because we wanted to carry around mini-PCs with us that happened to be used for voice communication.

    • Why would anyone want Nuclear to be flexible?…. It is simply a base load energy sources serving as a solid foundation to me base load requirements 24/7/365 days per year….and that is predictable until there is an overabundance of sporatic energy sources that can’t play nice with the existing baseload system….

      And that is just plain old bad engineering……an overly complex electrical grid system that becomes inherently unstable and too difficult to maintain and repair….. big trouble ahead IMHO.

      Now in the case of Germany where there is an abundance of electric trains operating mostly in daylight hours when wind and solar tend to work best in daylight hours…. that is a good match and there is no “24 hour duck curve”….

      But then again Germany lacks sufficient wind and clear skies for optimal solar and wind power…… Big gamble IMHO.

  3. As an electrical engineer with 35+ years of experience on both the electrical utility and client side….I use a $30,000 software program called ETAP so as to dynamically test out various power balance scenarios using fixed base load power generation in conjunction with variable alternative energy solutions including wind and solar…..

    We are looking at another 10 to 15 years of VERY HARD WORK to get these various power systems to play nice together. I can tell you that the control room operators and many others working the the large electrical power utility generation control rooms are getting very stressed and will start burning out much like the overloaded air traffic control room operators.

    Most of that stress is the result of unpredictable wind power generation.

    The consensus amongst the electrical engineers is…. thanks for the extra work and income but we really would prefer to meet this challenge without being micromanaged by Luddites who seem to think that the laws of physics are negotiable….. God help us all!

  4. I was on CASIO’s web site recently too as I wanted to see how much curtailment was occurring via the “Curtailed and Non-Operational Generating Units” report: The amount of forced curtailments on the system was much greater on the April 27th by the way.

    I was a tad surprised to see that Ivanpah 1 and 2 were noted as being “forced” to shed 123 and 133 MW. When I checked the hourly CASIO record for April 30 I noticed that one of the Diablo Canyon reactors, Number 2, was being shut down and the unit status report indicated that the 1150 MW curtailment was planned.

    I wonder why PG&E’s contracted (via a long term PPA) output from the Ivanpah facility wasn’t affected like the SCE output was……… I am aware that SCE valued the output of the Ivanpah facility differently than PG&E back when the original PPA contracts were signed (as denoted in the TOD factors and the price per kWh of the output in the contracts). I wonder if SCE is forcing the operators of the facility to meet the letter of the contract…… sounds like something I felt we had to do contractually with a strategic suppliers 20+ years ago.

    I was checking CASIO’s website at the end of April as our PV system (6.12 kW STS rating, 5.22 kW CEC rating) had reached it’s peak daily output (37 kWh) for a few days at the end of the month when the sky was clear. The El Nino storms have been great for our water table and plant life, but the output of our little PV system has been a bit lower on average the last 6 months (vs the monthly average output of the system: n=10 years). Our output on the 30th was effected by a Tahoe like thunderstorms that rolled into my area late in the afternoon. Our daily output dropped by 1 kWh, not too bad, as the day before our output was down by 4 kWh as our panels got cleaned naturally (0.3” inch of rain in about an hour). It looks like the storms effected the output of the Solar Thermal facilities in the state a lot more then they effected our little system.

    Our electric rate schedule from PG&E was eliminated recently (E-7 TOU with 2 time periods: peak and off peak) and we are just getting used to our new one (E-6 TOU with three time periods and prices; peak, off peak, and partial peak). Yesterday was the first day our rate schedule had a Peak time charge for our kWh usage (between the hours of 1 pm and 7 pm). If I had known that we were going to be forced off our E-7 rate schedule I would of put some of our panels facing west- it’s kind of hard to change their tilt and direction now.

    By chance have you seen an updated report from CASIO on the expected fuel burn as we move from the 20% RES to the 33%RES. If memory serves me correctly one of the Scenario’s run by CASIO a few years back indicated that the fuel burn could be higher at the 33%RES vs the 20%RES. Having an El Nino year will certainly help increase the flexibility our the system to minimize the excess fuel burn to keep the grid up.

  5. The scale on the “duck curve” are a bit misleading. 1) It should include 0 to show that the effect really is relatively small, and 2) the scale should go up to the system peak load (in August). This shows the true scale of the issue. In addition, it would be helpful to put this in the overall WECC load profile. Note that CA is usually importing midday. This solar output just decreases the imports. This results in less generation at coal plants elsewhere from the least efficient plants. So because California is interconnected, this looks much less dire in a larger context.

    • Although the dip is small, the peak demand that occurs at 8 pm is still load that must be covered by some capacity. Additional solar does no good for serving this load thus solar now has an incremental capacity value of zero in Calif. This means you can’t retire conventional generation as you add more solar. Until this problem is corrected you can’t use solar to replace fossil generation capacity, i.e. can’t retire fossil plants.

      • The peak demand that occurs at 8 pm is non-summer peak demand of 25,000-32,000 Mw. The mid-summer peak demand is far greater – 35,000-45,000 Mw, and it happens at 4-6 pm when the sun is still shining. So your comment that “solar now has an incremental capacity value of zero in Calif.” is true in the spring but false in the summer. And for capacity values in California, summer is what counts.

      • Gene, In addition to David’s point about the spring not being the dominant period for generation capacity, you haven’t acknowledged that California is connected to other regions and can sell in the solar peak hours and buy in the evening (when loads are falling in states to the east, in the Mountain Time Zone).

  6. Nice analysis. As long as there is over production during sunlight hours followed by no solar at sundown, the total energy contribution by solar is quite limited. This must be overcome if the global CO2 problem is to be solved. This means that Tesla’s battery storage system must be made to work. I did a small presentation here in Texas on the benefits of a home microgrid. see The home microgrid is more likely to develop than grid connected storage.

    • Dr. Preston, thanks for linking your presentation, which is very informative (and wonderfully concise). I agree that small, distributed storage units could be used to “buffer” power from peak solar or excess renewable production and still “play nice” with the existing utility grid through proper inverter design and appropriate tariffs. Certainly an alternative to building utility-scale storage, lots of new capacity or betting that natural gas prices will stay near basic production and debt service costs forever

      • Looking back in my studies I see that in addition to a zero CO2 plan for ERCOT of 144,000 MW renewables plus a 50,000 MW battery with 330 hours of storage, I see there is another alternative plan I had developed that works for near 100% CO2 reduction. It is a 35,000 MW of new base loaded nuclear (IFR?) and 12,000 MW of Per Peterson’s nuclear plant with GT peaking. This is for a system peak demand of 71,000 MW.

        So let’s assume the renewables costs $4/watt (some external costs such as transmission are included) for 144,000 MW which is 576 billion dollars plus 6.6 trillion for the storage (400 $/kWh) for a total of about 7 trillion dollars.

        Compare that with the highest cost nuclear of $10/w for 47,000 MW of new nuclear (has GT peaking built in) capacity is about 500 billion dollars. The nuclear could cost a lot less and not likely cost more than this.

        You can see that storage cost is a fatal flaw of the renewables plan. To be economic we need the storage cost to be no more than about 200 billion for 50,000,000 kW times 330 hours or 16.5 billion kWh of storage. This is a storage cost of about $12 per kWh. Elon would need to drop his promised $100 per kWh battery cost by a factor of 8. Also the batteries need to last at least 20 years. This probably eliminates Li Ion batteries. Only flow batteries can last that long and no one has any real plans for them at less than $400 per kWh. There might be a breakthrough on flow battery cost. Until there is a flow battery storage technology that costs a lot less than $100 per kWh, wind and solar are not going to eliminate the need for nuclear power to solve the CO2 emissions problem.

        Gene Preston

  7. Hawaii has the advantage of really high electricity prices due to its current dependence on what is essentially diesel fuel and it’s relatively small scale compared to most mainland systems. Consequently, Hawaiian utilities can afford to do things that could end up lowering consumer costs.

    Better coordination across the Western Interconnection looks good on paper but implementation is going to be difficult, especially for storage. Moreover, in the increasingly likely event other states follow California’s lead by raising THEIR RPS targets, grid coordination won;t be nearly enough. Imagine for a moment that the evening ramp is 40,000 or more MW over three hours instead of 13,000?

  8. Do you know why the 2013 actual load in the CAISO graph is so different from the 2013 actual load in your graph? Are your definitions of net load different?

    • Hi Paul:

      Good question. I could not easily find the net load data behind the original duck chart. And I wanted to plot an apples-to-apples comparison across years. So I downloaded the CAISO “renewables watch” hourly data and defined “net load” as the sum of nuclear, thermal, imports, and hydro. In other words, production minus non-hydro renewables. I average across March 28-April 3 to smooth out weather/day-of-week variation. I get a little closer to the original 2013 profile if I look only at March 31. But departures from the original remain. Perhaps some of our blog readers have better insight here.

      Thanks for reading..

      • The “renewables watch” hourly data that you use includes generation used to cover losses. To see actual load (measured at the CAISO point of delivery to load-serving entities, typical 60-200 kV), you have to go to the CAISO’s OASIS web site. Then you can subtract the solar and wind generation from the “renewables watch” website to get a more precise measure of “net load.” That, and using just March 31 versus the week surrounding it, explains why the ISO shape is different.

  9. Meredith: On a related aspect of your good piece you might want to see my “Revisiting Regional Regulation of Public Utilities” in the December 1993, Vol. XXVII No. 4 issue of Journal of Economic Issues, pp. 1219-1239.

%d bloggers like this: