At the risk of slipping into another debate, there are a couple of things here worth commenting on.
There are basically four categories of power delivery:
- base load plants --- power generation that can't be adjusted quickly, and are economically optimized to run at one load level
- - load-following plants --- power generation that can be adjusted slowly based on overall trends
- peaking plants --- power generation that can be adjusted quickly to keep the grid in balance
- - variable renewable energy (VRE) --- power generation that delivers what happens to be available at any given time (with curtailment as an option in the event of an excess)
It helps to look at real graphs of power delivery. California's main grid authority (CAISO) is a great example since there is tons of data, and other aggregating sites publish historical CAISO data. Here's a week in September 2024 that I graphed last year from one of the aggregation sites:
The categories are basically visible to the eye, to some extent:
- Base load: Nuclear, Geothermal, Biomass, Biogas -- near-constant contribution to the grid
- Load-following and peaking: Large and small hydropower, natural gas, batteries (there's also Imports which is power imported from neighboring geographies, but unfortunately this doesn't tell you what the fuel mix is within imports) --- hard to distinguish the two categories, but here's what Wikipedia says about the speed of
dispatchable energy generation:
The fastest plants to dispatch are
grid batteries which can dispatch in milliseconds.
Hydroelectric power plants can often dispatch in tens of seconds to minutes, and
natural gas power plants can generally dispatch in tens of minutes.
- VRE: Wind and solar
(There are no coal plants left in California; the last one was decommissioned in 2014.)
With that context:
what do you think about nuclear power? seems to be the only high density power source. we are generations ahead now in technology to make it safe and handle the waste.
Nuclear power is a base load. It's not going to help with balancing in cases of rapid load increase/decrease. There are good reasons to have nuclear generation (in my opinion), and if it can be made more safe and less expensive, we ought to displace some of the natural gas plants in the US to reduce our dependency on hydrocarbon fuels and to reduce CO2 emissions. That's my opinion; there are a wide variety of opinions and in this country we're not likely to settle the debate anytime soon.
Actually, peaker plants exist so you can build base-load generation to some calculated level, like the 95th percentile load, rather than the 100th percentile load. For generation agencies and companies, this reduces the cost of the base load plants, and saves fuel costs for fossil fuel plants because peaker plants can quickly started and shut off, while base load plants usually run at one speed (redline) and can't be easily or quickly shutdown and restarted.
That captures some of the distinction, but misses a few nuances.
The 95th/100th percentile load argument is wrong here, because grid usage is so variable on different timescales. In summer months in the US Southwest, because of air conditioning load, the ratio in electrical demand between peak demand (late afternoon / early evening) and minimum demand (pre-dawn) in a given day can vary by a factor of 2 or more depending on the area. But it's also largely predictable because of good weather forecasting* and demand modeling, so the CAISO energy markets anticipate it (there's a day-ahead market in hourly blocks, and a real-time market with 15-minute and 5-minute dispatching) and schedule the up/down patterns largely ahead of time. Because of this variability, base load plants can't provide 95% of the load
unless there is a place to put the excess energy. Batteries and pumped hydro storage (or even regular hydro, to some extent, by reducing flow) can take care of some of this energy on a minute-to-minute and day-to-day basis, but there is no technology that allows us to take large amounts of electrical energy and store it for long periods of time. So the "calculated level" here is where it makes sense economically to operate a plant for a long time or shut it down. Once you build a base load plant, because of the high cost of capital and low cost of fuel, you're basically going to operate it at this level forever, otherwise the economics don't work**. (Same with semiconductor fabs -- there's some variability, but if the utilization drops then it's no longer economically viable.)
In regions with
market-based grid operation (in the US, this is CAISO in California, PJM in the mid-Atlantic states / WV / parts of the Midwest and Kentucky, ISO-NE for New England, NYISO for New York, SPP for the Plains states, and ERCOT for Texas), peaker plants still exist because they are cheap enough and fast enough to remain economically viable even if you use them as infrequent reserve. They are not economically competitive enough to operate for longer periods of time, otherwise they would do so. Whereas non-peaking natural gas plants are optimized for economic efficiency over a particular utilization range, and are not designed to spin up/down quickly. In CAISO territory,
battery storage has grown phenomenally in the last five years, from near-zero in 2019 to 13,000 MW / 47,300 MWh in December 2024, with no signs of slowing; I would be willing to bet that natural gas plants are going to be the chief losers in California over the next few years, with many of them ceasing to operate because of displacement by battery grid storage. The peaking natural gas plants will die --- good riddance. Where you're going to see the need for natural gas is to handle the variability of power demand over the longer durations (weeks/months, with changes in weather patterns) because no other power source can provide that long-term adjustability; hydro gives you some seasonal variability, but the capacity isn't adjustable. But then we'll end up with a really weird situation where some natural gas plants operate only in summer and not at all in the wintertime, with variable needs in spring/fall, and that seems like a tough pill for plant operators to swallow. Or the solar/wind operators may have extended periods of curtailment... I don't really know what's going to happen.
In other regions where grid operators aren't fully deregulated (in the US, it's the western states other than California, and the southeastern states***; I live in AZ so I'm in this bucket), the mix of suppliers is dictated by policy and planning and the need to import/export power to neighboring regions for short-term minute-by-minute load balancing.
All of this discussion neglects one key resource: the grid's transmission capability. In CAISO territory,
this is captured by real-time location-based pricing that takes into account local grid congestion. In certain areas, notably near wind farms, there are times where the marginal cost of electricity will swing negative --- not because there's too much power in California as a whole, but because there isn't enough grid capacity to take all the power from that place and deliver it to another. Effectively the grid operator tells these wind farms, you can generate the electricity we scheduled yesterday, but if there's an unexpected increase in wind and you want to generate more power, we really don't want you do to that, and if you do, you'll have to pay us to take that power.
Grid transmission constraint is another huge reason that battery storage is competitive, since it can handle those mismatches in power in the short term (1-4 hour range) and reuse the energy later in time, to relieve congestion.
You might even find that data centers are building large banks batteries on-site, to handle their own peak loads.
If only this was the full story. The fundamental problem for the grid is that batteries take at least four hours to recharge to get another four hours of discharge. In the event of long lasting peak events with little wind or sun it's lights out.
Yeah, sort of, not exactly. Batteries are limited by technological reasons; charging/discharging is expressed in terms of "C", which is a one-hour recharge or discharge. So if you charge a battery with a four hour rate --- suppose you have a grid-scale battery with 10MWh and you charge it with 2.5MW --- that's C/4; a ten-hour recharge would be C/10, and a 15-minute recharge would be 4C. (Fast chargers for phones / laptops / electric cars allegedly get you to around 2C;
BYD claims 10C for some of its electric vehicles, but that's a lot of wear on the battery, and I wouldn't want to put my car through that kind of surge.) There's "unused" space at the top and bottom of the battery storage that is left to enhance the useful life of the battery, but otherwise, batteries can be charged as fast as their technology and ambient temperature allow, and economics permit. C/4 is not particularly stressful for batteries. I don't know what a typical grid-scale battery farm uses as a design/operating limit, but I would guess that C/2 is no problem and maybe even 1C, to take advantage of cheap electrical power during periods of excess in advance of expected periods of high demand. (Make hay while the sun shines!)
But you're right that batteries can't cover long-lasting increases or decreases in demand. (These aren't peak events, though.)
TL;DR -- it's complicated; nuclear power generation can't adjust quickly**; batteries are great for meeting short-term supply/demand mismatch and will probably drive out much of the peaking natural gas plants, at least where the economy of scale permits them to do so.
*
good weather forecasting --- gee, you'd think someone would realize this and not plan to
dismantle the National Center for Atmospheric Research.
**
nuclear power can't adjust quickly; once you build a base load plant, because of the high cost of capital and low cost of fuel, you're basically going to operate it at this level forever, otherwise the economics don't work --- well, it looks like in some European countries they are using
variable generation in nuclear power for load-following. edit: oh, here's the kicker:
Although this does not prohibit power load variations controlled by the operator (if justified from the technical and economic viewpoints), manoeuvring in automatic mode is not authorised by current regulations in the United States"
***
southeastern states: Looking at the
FERC page, it mentions the Southeast is a "bilateral market" which means each generator G1 has to find consumer buyers C1/C2/C3/... to take its electricity during each segment of the day, but that's not the same as a true auction market.
Disclaimer: I'm not a professional in the area of electric power, but I am an electrical engineer and I have been learning in-depth about the grid for a couple of years just due to curiosity.
)en.wikipedia.org
(Chose CNN specifically for this topic because I know you love CNN