I don't think the author understands the electricity market. Negative spot prices for energy are not unique to Texas, or wind.
Every energy production mechanism has a limited ability to scale production (either up or down) on demand. If you overproduce electricity, it needs somewhere to go, and the electric company is willing to pay consumers for the service of dumping the excess electricity.
This is not a market inefficiency, it is a technical inefficiency, and is essentially inherent to the problem of power generation.
If the cost was negative for a prolonged period of time, you would have a real story.
Worked on a project to address the problem of predicting wind power production over the short term (as in the next couple of hours). It is a very hard problem, and when I left it we had declared it successful because we were something like 10% better than the current method, which is just to assume that current production will continue.
Wind power is a strange beast. A 300MW wind farm can go from 250MW to 0MW to 100MW over a 1-2 hour period. Trying to properly plan for that or ramping up/down your coal/gas plants to cover the changes is very difficult and expensive.
I'm inclined to think that wind will only work at bigger scales if we can find industrial processes that are dominated by energy costs rather than capital costs and which can scale up and down in energy consumption fast enough to track the wind output.
I can think of a couple of possibilities. Water desalination is an obvious candidate. It might be economical to massively overbuild desalination plants and then only run them at peak capacity when wind is at peak as well. If the cost of electricity is much greater than the capital cost of the desalination plants, then it might be a net win when amortized over the lifetime of the plants.
Generating hydrogen by electrolysis and then immediately converting it to ammonia by the Haber process might be another. The production of ammonia for use in fertilizer using methane as the hydrogen source is a major green house gas contributor. There would be a significant advantage if we could move to carbon-free production.
The usual candidate for this is aluminium smelters.
They are extremely energy intensive: usually operators will only turn them on full when electricity is below e.g. $50 a MW.
The have the added bonus that they can act as a temporary battery: you can smelt extra aluminium to use up the negative energy rates, but then let it convert back again.
I thought aluminum smelters couldn't really shutdown production.
I remember it was big news in New Zealand when Tiwai Point (which uses 15% of the power nation wide) shut down some of it's production during the 2008-2009 power shortage (caused by a drought in the hydro power lakes.
An unplanned, no warning shutdown of a smelter would be bad - it ruins the rather large and expensive carbon electrodes used to melt the aluminium. Probably not so good for the crucibles and other equipment that might get frozen up.
Planned shutdowns are another matter entirely.
In terms of instantaneous interruptible load, wood pulping mills are king. There are some enormous wood chippers in NZ's north island that can be remotely turned on and off by one of the generating companies.
Also, just about everyone's hot water cylinder in New Zealand is remote controlled by the power company (and has been since the 50's) Allows instant shedding of quite a lot of load in the short term without any visible impact to customers.
In Europe I feel most talk is still on connecting Norway to say Germany with new electricity lines to utilise Norway's natural pumped storage potential.
Another interesting development recently was the creation of electric diesel. The economics don't seem to make a lot of sense yet but I don't think they've hit the efficiency limits of what is possible yet. The interesting thing is that it's carbon neutral (assuming the energy going into the process is carbon neutral, e.g. when excess renewable energy drives energy prices negative at moments of the day) probably works on a massive scale, and immediately allows existing fleets of combustion engines to become more sustainable. I'm not sure how fast these factories can turn on and off production but I'm assuming it's pretty good.
The efficiency is pretty low, electricity to diesel I've heard of efficiencies of 13%. And then a diesel engine itself is only about 30% efficient, compared to electric engines of > 90%. It makes more sense to invest in storage solutions and an electric car fleet, but if we do get significant excess renewable energy that is difficult to store, combined with ediesel becoming more efficient, then it becomes an interesting idea. Particularly for all the transport that is hard to electrify (airplanes, trucks), carbon neutral synthetic fuel made with renewable electricity could make sense one day.
I'm mostly optimistic about electrifying cars though. The next-gen top electric cars are getting close to ~100 kwh storage (today it's more like 80-90). Cars per capita in OECD countries is probably around 0.6 - 0.7 or so. That's a lot of storage if fully electrified, about 60-70 kwh per capita. Daily residential electricity usage is probably about 10 kwh in OECD countries. Of course that's when you electrify the entire car fleet which is a ridiculously big task (today's new car sales tend to be sub 1% electric), but if you can do 5% within a few decades, you can create multiple hours of storage which would be a huge impact on the ability to balance intermittent energy sources, without having to invest in storage for the sake of storage.
I consider home battery packs a much greater sign of market dysfunction than negative spot prices. Notably Tesla are targeting Hawaii and Australia as early markets. The grid should be distributing that energy at a fair market price, if it makes economic sense for an individual to avoid the grid for that service then there's probably an issue somewhere.
What's the difference between market dysfunction and market disruption?
It seems to me that home battery packs make things like wind power much more attractive because of their ability to smooth out the difference between supply and demand.
I'm not sure why that is dysfunctional. It seems to me like it is compensating for a technical issue?
It's dysfunctional because the balancing of supply and demand can be done cheaper and more effectively as part of the grid. But the grid operators often have poorly aligned incentives and so don't bother to do this.
When looked at from a holistic view, the grid should pay much more for distributed solar production. If it was valued correctly then consumer battery storage would rarely, if ever, make economic sense.
Hawaii have been refusing to allow more solar to be added to the grid, instead of investing on the network upgrades to make it work. Australia are charging high electricity prices for overbuilding the network and not properly valueing the contribution of solar. Various other places are trying to introduce punitive charges for domestic solar. It's good that consumer storage exists as it puts market pressure on these dysfunctional organisations, but decent regulation would have sorted this out properly with far greater efficiency.
I suppose in Hawaii the "issues" are the whole "predictable and plentiful sunlight for about 12 hours a day all year" combined with the whole "Pacific Ocean" thing. Solar could easily and cheaply provide all of Hawaii's electricity needs if there were a good way to store the energy for overnight usage. Tesla's home battery packs are actually a great interim solution for such a market, until some larger scale instant-on grid-wide solar energy storage solution is available.
Battery packs at the end of a link are not a bad plan though, as they bring down the peak required on the link and so can actually be cheaper than a fatter cable.
This should change with electric cars. I'd love to have an electric car set up to recharge when the rates were low. This could really even out the power consumption.
Heck, I can think of a lot of uses of power that become practical if the rates are really low. The hot water heater could be heated up a bit extra hot. Ice could be made to supplement the air conditioner during high electric cost times, etc.
All that's needed is variable pricing to the consumer.
It can be set to change when the price is cheapest (on a short timescale, not just an off-peak price) and also to try to minimise the carbon used. It does depend on the local utility offering rebates, but it's often in their narrow financial interests to do so, so the incentives work out.
See also Ice Bear (and competitors) for air-con that stores cheap electricity as ice.
This generally falls under "smart grid" and "demand management" and well designed electric grid markets like in Texas enable these companies to respond to the market price signals.
Yup. I seriously do not understand why electric utilities do not do variable pricing to consumers. Do it, and the 'problem' of electric surpluses will go away.
There's usually some cost differences for night/day time (or winter/summer) depending on your contract
Now, to have variable pricing for home customers I think the complexity outweighs the advantages (as in, limited capacity to enjoy momentarily cheaper prices)
Usually how this is done is that you get a cheaper tariff and give up some level of control, e.g your air-con might be shut off for 5 mintues (along with 100 of your neighbours) to respond to a demand spike elsewhere.
The "power company can turn off your AC" thing never made any sense to me at all. Why not just let the customer set the price at which they will automatically turn off their AC, and let the power company know what it is. Then you get automatic dutch auctions for the price of electricity that set the price where demand meets supply.
Then you can set different prices for different things: Don't turn off my AC unless the price exceeds $.50/KWh but stop charging my electric car if it exceeds $.12/KWh.
Because "power company can turn off your AC" is not a feature for you, it's a feature for the power company.
The power company's problem is that everybody has their AC turned on at 3pm and either the node supplying those houses is overloaded or they must buy energy at crazy spot prices to clear the market.
If they can flip a switch and knock out nearby ACs for 10 minutes then they have time to figure out how to solve the problem before it becomes a blackout.
Now they could "theoretically" introduce a pricing mechanism such that a number of people would shut off their AC due to an increase in price. But that is a lot more complicated (most residences do not do real-time billing for energy, harder to forecast how many users will shut off if price is raised, have to maintain real-time channel for price data) and in practice something will go wrong and you'll blackout instead.
> Now they could "theoretically" introduce a pricing mechanism such that a number of people would shut off their AC due to an increase in price. But that is a lot more complicated (most residences do not do real-time billing for energy, harder to forecast how many users will shut off if price is raised, have to maintain real-time channel for price data) and in practice something will go wrong and you'll blackout instead.
Doing real-time billing is not a hard technical problem, it's just a one-time cost to replace all the meters.
At that point all you actually need is a standard IP-based protocol to announce real-time rates. The power company doesn't even need to communicate bidirectionally with individual devices whatsoever -- all they need to do is broadcast the current rate and let the devices make choices. Because the choices various devices make will invariably be "when price goes up too much, use less power." So if people are using too much power, you raise the rate a little bit at a time until they no longer are (or the higher rate causes more expensive generation methods to come online).
The only way you get blackouts is if the rate goes so high that every smart device with a cutout is already off and you still don't have enough power, but that's exactly the same failure point as letting the power company turn off your AC. Only worse because if you need some non-standard power company interface with bureaucratic rules instead of a simple standardized rate announcement protocol then there will be fewer smart devices built and purchased that will reduce consumption in response to high rates.
> it's just a one-time cost to replace all the meters
Much easier said than done.
> At that point all you actually need is a standard IP-based protocol to announce real-time rates. The power company doesn't even need to communicate bidirectionally with individual devices whatsoever -- all they need to do is broadcast the current rate and let the devices make choices
Several problems there (before I get into 'kids these days' mode):
- Meter connected 24/7 just to get prices is not feasible (also remember you need per meter pricing for the issues
listed by the parent)
- For it to make decisions to "use less power" you need a controller for every device (heating/AC/etc) you want to be controllable - or something more complex if you want fine-grained control
Too complex for home usage, basically, where you want to turn your devices on and not worry about spot pricing
Said chargers simply wait until midnight to begin charging, when power is at its cheapest.
My power in Northern Illinois from nuclear plants run by Exelon is 1 cent per kwh between midnight and 5am due to such low demand. That's pretty close to free.
But I'd sign up for variable pricing. I'd need a way to query the internet for the current price. Then, appliances can be set to run when rates are the cheapest. (Electric dryer, hot water heater, refrigerator, dishwasher, car battery charger, etc.)
If they're paying people to dump electricity on them, I can't believe they've got variable pricing for the bulk of their customers.
Not really. Most people have pretty inflexible schedules, which dictates when they are home and when they are sleeping, etc. and thus when they are using electricity.
Though this evens out a lot when you have many farms with some distance between them.
Many coal and nuclear plants are actually quite good at load following (at least in France and Germany it's common to run nukes in this mode). The units are quite big so they don't have to do that many % per hour in order to keep up with a wind park.
"Though this evens out a lot when you have many farms with some distance between them."
Unfortunately, it doesn't. The big grids such as CAISO (all of California, with wind farms over a range of 500 miles) and PJM (Ohio to the Atlantic Ocean) still routinely see 4:1 variations over the course of a day.
Here's today's CAISO data: Max wind generation: 1350 MW, at 1 AM. Min wind generation: 186MW, at 5 PM.[1] This sucks. At least with solar, you get max output and max air conditioning load at the same time. Getting max wind generation at 1 AM is not too helpful. However, the California Water Project, which is a big power buyer for pumping water uphill, profitably buys late-night cheap power for pumping, then lets water back down through turbines during high-load periods. So some of the big reservoirs double as pumped storage.
Batteries would help. It will take a lot of them. At least 300,000 Tesla Powerwall battery packs would be required to load balance that variation in wind. That's about a billion dollars, which is not unreasonable. When we see wind farms starting to buy batteries in bulk, it will be clear this is working.
But the amount of battery required for reliability is higher than the amount required for improved profitability. Something has to be able to provide power when the wind is calm for a few days, which happens.
Variation and unpredictability are two different things. It varies by geography, but often wind is stronger at night and in the winter but that is different from minute by minute variation, which is what can be reduced by building lots of distributed wind farms.
Solar, for example, could be said to have infinite variability since its not available at night, but its actually fairly predicatble if it is spread around geographically.
Batteries would help. It will take a lot of them. At least 300,000 Tesla Powerwall battery packs would be required to load balance that variation in wind. That's about a billion dollars, which is not unreasonable.
The price would of course go down for quantity, and putting the individual cells into larger units. Let's wave our hands and say 900M -> 500M.
But per https://en.wikipedia.org/wiki/Tesla_Powerwall the $3,000 7kWh daily cycle units have a lifetime of "only" 5,000 cycles. So half "a billion dollars" every 13.5 years, ~100K/day.
For air conditioning (heating or cooling) you don't need to go as high tech as batteries - you can heat/cool water during the night and use it in daytime. This is common in heating applications (due to longstanding night-time electricity discounts) in cooler places of the world. Also works for district cooling.
> Many coal and nuclear plants are actually quite good at load following (at least in France and Germany it's common to run nukes in this mode). The units are quite big so they don't have to do that many % per hour in order to keep up with a wind park.
It doesn't really make sense to do this with nuclear. Once you've paid the capital cost to construct a nuclear plant, the fuel cost is completely inconsequential and you might as well run it at 100% constantly. Even if you can vary the load you're better off to find something to use the extra capacity for than, essentially, throwing it away.
Of course, that's before politics and market inefficiency. It may be more profitable to reduce supply unnecessarily if it will raise the price per KWh because it's more profitable to generate 900MWh at $.18/KWh than 1000MWh at .04/KWh.
In the power system there are plants with different roles. Some of them will be taking care of balancing the power system and others will be mostly running at capacity. Both tasks provide value. You don't need any "unnecessary" reduction in the equation if it's cheaper to slow down some plants than feed the excess power into the market at negative prices.
I have watched and read about this topic before and got the impression that it is essentially a solved problem with a prediction accuracy of better than 5 % and a time resolution of 15 minutes one day into the future. Is this not good enough? Are the prediction not actually that good?
While this is true, I have a feeling that wind averaged over a big land mass like US, Canada and Alaska will be a very close to constant. What we need is a better buffer system - here is where the supercaps, once developed could play a huge role. Or hydrogen/ fuel cells.
This reminds me of a 1970's Boy Scout visit to the local power plant. It had a coal and a gas fired boiler, and the tour guide mentioned that that night, the gas company had asked them to preferentially use gas, as I recall the guy said the pressure in the pipelines was getting high.
This happens occasionally in the PJM control area, even without a subsidy. PJM has a lot of nuclear and hydro plants in their territory. The nuclear plants don't like to cut output at 4 AM when they'll have to crank it up at 5:30 AM; the thermal time constants in the system are too big. So they'll take an occasional hour of negative prices.
Hydro Quebec sometimes has too much water, in which case they run all their generators at capacity. They, too, will take an hour of negative prices rather than closing huge steel intake gates and taking a generator out of service for an hour, only to have to go through the startup and resynch process an hour later.
Wind output in the PJM area varies about 4:1 over a typical day. When peak wind lines up with minimum power demand, prices go negative, the automatic bidding system communicates back to the wind turbines, and the blades go to zero pitch, slow, and stop. The wind guys grumble about this, because some of them financed their operations without allowing for this. But it's not frequent. Wind turbines are good at shutdown and startup resynch, because they have to do it every time the wind slows down below useful speed.
Interesting topic; silly article. The author really wanted to use the "only in Texas" tag, but if you read closely you notice the main driving factor was a Federal subsidy; nothing at all to do with Texas. (And as other commenters have noted, other regions and countries also have occasional negative spot prices.)
I'd say something snarky about "only in slate", but of course, they're hardly unique.
So, Texas constructed a market that allows it to take advantage of Federal dollars being credited to other market participants (energy providers) in the most efficient way possible.
The author didn't clarify the "complicated" market, but my assumption of the highest-price-to-everyone mechanism is that ERCOT only selects a limited subgroup of all bidders each auction period, according to whatever is most economical for ERCOT. So there is some incentive to keeping your bid low.
What I don't understand is how anyone who isn't benefiting from the tax credits could bid at or below zero. The author seems to indicate that wind did not meet 100% of the need during the negative price period and did not indicate anyone else was receiving tax credits beyond wind producers. So how was the highest price in the negatives? Someone had to have costs associated with the electricity they were producing within the 70% that wasn't wind-powered.
>What I don't understand is how anyone who isn't benefiting from the tax credits could bid at or below zero.
It's also a pretty cheap way of advertising a market opportunity and driving investors' expectations: your money would be better spent on a new aluminum smelting plant being our customers rather than a new coal or natural gas plant being our competitors.
Believe it or not, in the early days, the oil industry also had this problem of "excess supply". In the UK in the 1900s they dealt with this by lobbying Winston Churchill (before he was prime minister) to shift the navy from running largely on coal to running on oil.
It's not the highest price bid. It's the highest price bid among the cheapest set of bids which could fill demand. Say we need 100 units of electricity in the next hour, and the bids are 50 units @ $2/unit, 30 units at $5, 30 units at $10, and 30 units at $20. Then, to get 100 units, you only need to buy from the first three suppliers, who are offering 110 units between them. The highest price among those is $10/unit. The 4th supplier's price is higher, but he's not among the winning bidders.
That's what I'm saying. Let me rephrase within your example: How was the highest price bid among the winning bidders negative?
According to the article, not all of the electricity could be provided by wind. So roughly 70% of the capacity offered by the "winning bidders" came from non-wind sources which, presumably, require positive prices to sustain themselves. How was the highest bid among those selected negative?
As noted in the article, wind power is highly variable, but other sources have the opposite problem: Slow ramp up and ramp down times.
If you're running a massive coal plant (and nuclear is ten times worse), you can't just turn it off for 5 minutes if the last 5 minute auction didn't go your way. Even if you think the next 10-20 auctions won't be economical for you to win and you start the shut down process, you'll still need to get rid of that power you're generating in the meantime.
What (probably) happened is that the a combination of high wind production and low demand meant the system temporarily had a surplus of power, and some (non-wind) producers actually were selling power at a loss just to get rid of it because they couldn't wind their systems down fast enough. Which is a much more interesting story than "federal government subsidies wind". Shame Slate didn't talk about it.
Your point is an interesting one and leads me to an odd conclusion: the presence of wind producers in this market is not a requirement for negative prices.
Assuming the author is accurately describing the market, by definition, the bids of wind producers will always account for a minority of the capacity. Which means the other 70+ percent of the supply is coming from other producers. If your explanation holds, then coal producers and the like were bidding themselves at a loss, which seems like that is a state that could happen even in the absence of wind producers.
I can see how the wind producers are affecting this, but they're not a requirement. If there was a transition to considerably lower-usage electrical devices within Texas, the relative gap between demand and capacity of the existing producers would result in the same effect on the market: everyone scrambling to deal with ramp downs even at a loss. It's just that the wind farms produced that gap on the supply side instead.
no wind is not a requirement, but the variability wind adds is so strong this may never happen in practice without wind power.
if there was a general move to lower power use, the market would adjust to the lower level. it is the unpredictable variability that causes the negative prices.
Others have pointed out that other plants (nuclear especially) can't shift output quickly, so they'd rather pay to dump energy for a few hours than ramp down and ramp up again.
I wonder if some producers were counting on the "winner's curse". The auction pays out at the highest price that was needed, so if you want to sell at any price you might put in a negative price just to ensure you're picked.
As someone who interned at the New York ISO, the organization that operates the bulk electricity market in New York State, this article doesn't seem well researched:
> Third, Texas has a unique market structure....
The article then goes on to describe a market structure that is very similar to New York ISO and ISO-New England. It is called Location Based Marginal Pricing.
PJM has most of their training materials on line.[1] Those give a sense of how PJM does it. Start with "PJM 101", "Generation Basics", and "How PJM operates and dispatches".
PJM has a day-ahead market, where most of the capacity is booked a day in advance, and a real-time market, where differences from the predicted load are handled. This allows everybody involved to plan ahead a day. There's also "non-economic operation", where, in the event of big problems, the PJM control center just tells generators what to do. That's for emergencies.
All generators post not just a single price, but a price/output curve. Generators also post a "ramp rate"; how fast they can ramp output up or down on request. PJM's job is to optimize all this, which happens on about a five minute cycle. There are also lots of physical constraints - transmission lines and substations can only handle so much power, and policy is to run the system so that any single failure won't take it down. After a failure, the transmission system is reconfigured to become single-point failure tolerant again. There's a control room in Valley Forge, PA, (and a backup control room somewhere else) where about ten people control the wholesale power grid.
California had, briefly, a system with an spot power auction every half hour, with dealers who weren't generators or power consumers, but that was abused and produced blackouts, along with the bankruptcy of PG&E. California now has a day-ahead market and a spot market for fine tuning, like PJM, which resulted in much saner operation. So does ERCOT, the Texas grid manager. It's not a raw auction every half hour.
As an aside, it annoys me when people straw-man economists. It is of little consequence, but obviously any economist who studied the situation would understand at once why producers are incentivized to do this.
This is happening occasionally here in germany[1], most recently 2 weeks ago[2]. Here it's not just oversupply but also the lack of overland lines to distribute wind power from the north to the south, so we have to push it to neighboring countries instead.
Could this price make it to ordinary consumers? Could a consumer who was monitoring the prices consume a large quantity of power during the period to drive their on electricity bill down?
I live in Texas and I've never seen a plan more complex than a couple of tiers of prices depending on consumption, or different prices depending on the time of day (e.g. nights and weekends free).
So no I don't think any residential customers have access to electricity priced in any quantity besides a month of usage.
So this isn't a free market. This negative price is the result of an artificial market created by government subsidies.
If someone is selling something at negative prices, then there is a business opportunity for someone to buy that product and throw it away. Someone could setup a giant space heater to do nothing more than absorb electricity. The windfarms would pay this entity to keep buying power at no cost until the price approached the subsidy. Perhaps a steel mill or aluminum smelter could be paid to fire up another furnace for no other reason than to suck up power. I'm all for free markets and green energy subsidies, but I am totally against waste.
Hopefully some market actor will appear with the ability to store and resell energy. That might be homes with batteries and net metering.
No. The article states that electricity is sold at the highest bid to meet the demand, and that wind was only supplying up to 40% of the demand. I.e. It is the coal and nuclear power plants that are also bidding negative.
The market is working, to such an extend that during times of extreme oversupply power generators have to assess if they will temporarily turn down production (which has a cost, especially for coal/nuclear, as they are difficult to switch), or if instead they are willing to pay people to take the electricity.
Yes, without subsidies, wind would probably minimum bid at a $0/MWh, but that would still leave fuel based producers running at negative marginal cost.
What's missing in this market is flexible demand. I'm pretty sure if our appliances would spin up real time at low prices that it would be almost impossible to hit negative prices again, as demand will rise much more rapidly in response to lowering electricity prices.
Your giant space heater would make a tiny amount of money for an hour or so in the middle of the night, every once in a blue moon when a dip in demand and a spike in supply happen to overlap, taking advantage of the fact that the other (oil, hydro and nuclear) power plants can't power down and back up at such short notice.
If this happens all the time, it will drive the average price of power down, and power hungry industries will move there, but steel mills and aluminum smelters can't move around the country at a few minutes notice.
Presumably in a wind turbine you can adjust the pitch of the blades and bring the rotor to a stop.
I agree that in the case of a nuclear power plant it's hard to stop generating power if the price goes negative - but if there were no subsidy presumably wind generators could switch off easily if they wanted to.
At the scale we're talking about it's not trivial - the sheer amount of energy you're dumping has to go somewhere, probably turning into heat one way or the other.
It would be interesting if this could be linked with hydro, and use -ve price points to pump water into reservoirs that can release power a short notice via hydro within a few minutes of demand. Then refill at -ve price points. Using wind and gravity to make money.
Germany with its switch to renewables is seeing the same issues. They pay the Netherlands and other neighbors to dump German surplus electricity in peak wind hours.
Germany has a goal of 80% energey out of renewables by 2050, but it is entirely unclear if this is even physically possible because of this issue of spikes.
Norway has it easier as it mountaineous enough to allow hydro everywhere.
Completely inaccurate headline, and they buried the lede. The real reason power producers were willing to 'sell' electricity for negative $8 per MWh, is because of a $23 federal tax credit they received per MWh. So the negative sale wasn't negative at all.
"Impossible, most economists would say. In any market—and especially in a state devoted to the free market, like Texas—makers won’t provide a product or service at a negative cost. Yet this could only have happened in Texas, which (not surprisingly) has carved out its own unique approach to electricity."
I sometimes wonder if it is a point of pride for journalists to be ignorant of the subject of their articles.
edit - also, in a regulated market, one of the things that is very commonly regulated is companies offering goods or services below cost, as it is a very common tactic for manipulating markets or driving your competitors out of business.
Every energy production mechanism has a limited ability to scale production (either up or down) on demand. If you overproduce electricity, it needs somewhere to go, and the electric company is willing to pay consumers for the service of dumping the excess electricity.
This is not a market inefficiency, it is a technical inefficiency, and is essentially inherent to the problem of power generation.
If the cost was negative for a prolonged period of time, you would have a real story.