Today’s energy experts are increasingly questioning the validity of the generations-old way of thinking about our electric grid that says coal is needed as a baseload power resource year-round. “Baseload resources” are the generators that are thought of as providing a constant stream of electric power year-round. At its simplest level, energy wonks tend to fall in one of two camps:
- Those that argue baseload resources like coal and nuclear are needed to reliably provide a constant amount of electricity year-round and so have to be preserved through additional incentives.
- Those that push back on the very notion of the need for baseload resources, noting the way we think about providing electricity should recognize the favorable economics and improved performance of technologies like wind, solar, and batteries.
This blog attempts to illustrate that:
The very notion that “coal is a baseload resource” is wrong. Plus, it turns out that not running a coal plant as a baseload resource can save customers 10’s of millions of dollars.
History of baseload
Historically, electric demand has been thought of as being either baseload, shoulder load, or peak load.
Baseload demand is the minimum level of demand that the system in aggregate never (or rarely) falls below. Peak load is intuitively the load associated with the peak in demand, the highest levels of demand that are only seen during the highest demand hours of the day (daily peak demand) or highest days of the year (annual peak demand). Shoulder load is the demand that falls in-between the peak and base.
For the first 100 years (or so) of the utility industry, baseload resources were built and operated to serve baseload demand, that is, they were built to turn on and stay on, running all out all year round and meet the portion of demand that appears constant in aggregate. Baseload units were built with the expectation of having a high capacity factor and have historically had high capital costs but low operating costs.
It is worth noting that wind and solar have high capital costs but low (nearly zero) operating costs.
Shoulder units were used to serve shoulder load (also known as “mid-merit” because they fall in the middle section of a merit order supply curve. Shoulder units or mid-merit units have typically been built on the expectation of moderate capacity factors.
Finally “peaking units” or “peakers” are typically built to serve peak load. Peakers tend to have low capital costs but higher operating costs compared to the mid-merit and baseload resources.
Traditional thinking was that market prices follow load, thus baseload, mid-merit, and peakers all have a corresponding economic meaning. The idea was baseload units are economic to run in all (or nearly all) hours of the year. While peaker units only make economic sense to serve the peak loads.
In the above illustrative graph, the system load (represented by the load duration curve) never drops below 20,000 MW and so we can think of the first tranche 20 GW worth of resources as the resources that are truly baseload when lined up in merit order. Next, you have a tranche of shoulder resources that fall in the middle of the merit order, i.e. are “mid-merit” resources. Finally, you have the most expensive resources, presumably only needed when demand is at its peak.
When wholesale markets were first formed, it made economic sense to continue running coal plants as baseload resources because the market clearing prices were, generally, above coal power plant’s production costs, absent a carbon price like those established by RGGI or CA AB32.
Those units were early in the merit order line and fell far to the left of the supply curve. So, if it runs at maximum, all year round, its owner might actually make money. That dynamic has changed.
One measure of how it has changed is the dark spread. The dark spread is the measured difference between coal production costs (which is dominated by coal fuel prices) and market-clearing prices. The difference between gas production costs (which is dominated by gas fuel prices) and market prices is known as the “spark spread.” A positive dark spread (or spark spread) is indicative that a power plant with the corresponding fuel source is generating net revenues in the wholesale energy market such that it is could be covering its fixed costs and maybe even earning a profit. Maybe.
Changing market conditions have reduced dark spreads and, in some markets, the dark spread is negative for sustained periods of time. EIA assembled this data for PJM and found that while spark spreads have been narrowly positive, dark spreads have been negative for many days, weeks.
A number of factors have contributed to declining and flat market prices. Increasing gas capacity coupled with relatively low gas prices have driven this changing market dynamic along with flat demand (thanks to energy efficiency) and increasing adoption of zero marginal cost resources like wind and solar. These factors in aggregate have turned power plants that were formerly economic as baseload to units that may only be economic at limited times, as illustrated in the below figure.
The idea here is to re-examine the past assumption of baseload plants operating at full output all year if they are not always profitable. The profitability of operations can and should be assessed in useful and meaningful periods. A regulatory review need not accept operations that lose money part of the year. Because power plants can ramp up and ramp down, it is possible for a power plant owner to operate a power plant such that it makes money over the course of the year, despite running at a loss for long periods of time. For example, a power plant might be making money in the summer month and operate at full output. During the months when the market price drops below a unit’s production costs it isn’t uncommon to see that the unit’s owner operates the unit at the economic or physical minimum output. This helps minimize losses, but not fully avoid market losses.
Cycling and Seasonal Operations
It is worth noting that some power plants have reacted to the changing market dynamics. There are two emerging trends in coal plant operations: cycling and seasonal operations
Cycling is a technical term to describe power plant operations that includes ramping the output of the power plant up and down over the course of the day (or week) to respond to changing market prices. Though cycling a coal plant does increase maintenance costs, it has been part of the strategy the fleet of coal plants in the Midwest have taken up.
Based on data assembled by the Energy Information Agency (EIA), coal-fired power plants in the Southwest Power Pool (the RTO directly west of MISO that spans from North Dakota down to Oklahoma and spills into small sections of New Mexico, Texas, and Louisiana) cycle on a daily basis. This claim is clearly supported by the data which shows the hourly variation in output of coal plants over the course of the day.
Cycling alone is not proof positive that a power plant is operating economically. Coal plants could be cycling, but depending on market prices it is possible that no amount of cycling will help a coal plant become economic. However, many owners of coal plants that don’t cycle suggest that it is operationally impossible to cycle coal plants because they are “baseload” resources. But the existence of coal plant cycling is evidence that coal plants can cycle. The question we should be asking is, what about your coal plant is “special” such that it can’t cycle?
There is a second trend amongst coal plant operators in response to changed market conditions. These plant owners save customers money by switching to seasonal operation and operating less.
This strategy has been adopted by the owners of coal plants in multiple states. The Martin Lake and Monticello coal plants in Texas, owned by Luminant went to seasonal operations in 2015. The Gibbons Creek coal plant, also in Texas, owned by the Texas Municipal Power Agency and member utilities, followed suit in 2017. Cleco and SWEPCO joined in on the trend in 2018 when the companies announced that their Dolet Hills facility, in Louisiana, would switch to seasonal operation. Xcel Energy has also begun exploring options to go on a “coal holiday” during the months where it no longer makes economic sense to run their coal plants.
When the owners of the Dolet Hill coal-fired plant made the switch to seasonal operations, they opted to only operate four months of the year between June and September when demand (and electricity prices) are highest. The utilities’ own estimations indicate this will save customers $85 million by the end of 2020. When Xcel made the switch for their plants, they announced that the move would save customers upwards of $30 million a year and about 5 million metric tons annually.
The issue will only get pronounced in the future
It seems like hardly a month goes by without a report coming out that shows utilities can build wind or solar at prices below the operating costs of the coal fleet. So, even if this issue doesn’t apply to a coal plant today, it is likely to apply to it in the near future. Utilities need to be constantly evaluating this issue in processes like plant scheduling, utility resource plans and requests for fuel cost recovery. Some utilities do a very bad job at this, and just assume their coal plants will run all out, all year round, out into the future indefinitely. But that type of thinking has to stop.