Why CSP Should Not Try to be Coal

Spread the love

Joe Romm, at the influential Climate Progress blog, has hit on a formula for countering the coal industry’s claims that we need baseload power sources.  Since Concentrating Solar Power (CSP) in conjunction with thermal storage can be used to generate 24/7 or baseload power Joe has renamed it "Solar Baseload."   This is win-the-battle-lose-the-war thinking.  While it does neatly counter the argument we need coal or nuclear, since there are renewable power sources which can produce baseload, such as CSP, Geothermal, and Biomass.  I fell into this coal-industry trap myself in a 2007 article about Geothermal, as did AltEnergyStocks Editor Charles Morand in an article on CSP.

Dispatchable Solar, not Solar Baseload.

Continuous power from solar energy was first demonstrated at the Department of Energy’s (DOE) Solar Two project in the late 1990s. I recently interviewed Bill Gould, CTO of CSP company Solar Reserve.  Solar Reserve is now working to commercialize the molten salt thermal storage and solar receiver technology demonstrated at Solar Two, where Bill Gould served as project manager.  

According to Gould, DOE’s intent at the Solar Two project was to demonstrate dispatchable power, not baseload power.  Dispatchable power is power that can be called on when needed, in contrast to baseload power, which is essentially always on.  As a demonstration, Gould’s team throttled back the output from Solar Two to 10% of capacity, and this allowed the plant to produce power continuously for a couple weeks until it was interrupted by several consecutive days of cloudiness.  But, in essence, it was a stunt: baseload power is far less valuable than dispatchable power.

The coal industry says that we need baseload power because our refrigerators still come on in the middle of the night.  This is like saying we should have the water running constantly in the kitchen sink because we may get thirsty at any time and want a drink.  Put in these terms, the assertion that we need baseload power is clearly nuts: what we need is controllable power that’s there when we need it, but is not wasted when the lights are off and the fridge is not running.  

The Problem With Baseload

Last spring, I discussed one of the problems with baseload power.  The more baseload power you have, the harder it is to use variable generation such as photovolatic (PV) solar and wind power.  Or, from the baseload generator’s perspective, the more variable generation on the grid, the less baseload power can be added.    This fact has not been lost on the UK’s nuclear industry, which is fighting to get wind targets lowered.

To illustrate the incompatibility of baseload and variable energy sources, I downloaded 4 days of real demand data (January 1-4, 2008) from ERCOT’s website.  I then simulated production curves for two variable sources, one designed to mimic solar PV (only on during the day, with some variability due to clouds) and a more random type of generation to simulate wind.  I then fixed the amount of baseload power at 25,000 MW (68% of demand) and 5,000 MW (14% of demand) in each of two scenarios, and saw how much wind and PV the remaining demand could accommodate with the constraint that total generation could not exceed demand.



As you can see, when I dropped baseload power from 68% to 14% of demand, I was able to increase the power of variable sources from 10% to 36% of demand.  Almost half of the drop in baseload power was filled by variable power sources, with the balance requiring an increase in dispatchable generation.  If you’d like to try your own scenarios, you can download my Excel spreadsheet here.

Better than Baseload

It should be clear that  dispatchable generation is a truly premium power source.  Dispatchable generation, like energy storage, long distance transmission, and demand response, all allow the grid to accommodate more variation in both power supplies and in demand.  In a carbon-constrained world, where we want to use as much variable generation such as wind and PV as possible, zero carbon, dispatchable power from CSP can do far more to help us decarbonize the grid than CSP baseload.

Baseload power is part of the problem; it’s not the solution.  We should not denigrate CSP by pretending it is only a substitute for coal or nuclear.

Concentrating Solar Power is much better than baseload.

Tom Konrad, Ph.D.


  1. An interesting way to augment this analysis would be to attach costs (say average costs) to the baseload, solar PV, wind and dispatchable power, and see how total average cost changes as you reduce baseload and increase the share of renewables and dispatchable.
    I suspect costs would go up, which is one of the main rationales for having baseload in the first place (large fixed costs amortized over more output). This might provide one hint into how expensive carbon would have to be to justify, on economic grounds alone, moving to your scenario #2 (given that a majority of baseload is coal and nat gas.)

  2. There are a lot of ways that I could improve this… it is simply a demo to show that dispatchable is more valuable than baseload. Here are some:

    • Use real costs
    • Use wind and solar data
    • Use a full 8760 hour year worth of data.
    • Acknowledge that coal is actually about 75% baseload and 25% dispatchable.
    • Acknowledge that not all dispatchable power is alike… ramp-up and ramp down times vary greatly (and more
      efficient NG plants tend to be slower to ramp up and down.

    In other words, this study could easily become a Ph.D. thesis… the first thing to do is to see what others have done, a process I’ve been working on for a few months.

    However, in terms of costs, here are my mental approximations:

    FUEL /   Price/Terawatt

    • 1 TW Coal   =  1 planet+ $0.7 trillion
    • 1 TW N. Gas =  1 ecosystem + $1 trillion
    • 1 TW Nuclear  = $3 Trillion + 25% chance of 1 planet
    • 1 TW “Clean Coal” = 2 ecosystems + $4 Trillion
    • 1 TW Solar PV  = $4 Trillion
    • 1 TW Wind =  $1 Trillion
    • 1 TW CSP =  $3 Trillion

    This simplifies the cost calculations.  Try not to spend any planets;
    you’ve only got one.

  3. A number I’ve heard recently is that in the classic fossil fuel model, dispatchable power (especially the kind covering the last 1% of demand) is 100 times the cost of baseload power. No wonder the fossil fuel people don’t like CSP.

  4. Dear Tom,
    I think your terminology has a lot of merit. The main difference between baseload an dispatchables seems to be the cost of producing them, traditionally they have been coal or nuclear and gas – with reasonably well defined relative costs.
    The relation between baseload and dispatchables seems to be that baseload is always there. Power from baseload source could become a dispatchable if we stored it in a battery or pumped hydro. A dispatchable could become baseload, but in the recent past we have not done this because of the relatively higher costs of dispatchables.
    When we introduce intermittent sources of power we run into problems with our way of thinking. At face value intermittents can only be used when they are there. But with storage (batteries, steam, molten salt, pumped hydro) intermittents can become dispatchables. And dispatchables can of course become baseload.
    So in the new world we are facing we have intermittents and they can be transformed into baseload – at a cost.
    As you rightly show in your diagrams there are multiple mixes of baseload, intermittents and dispatchables that can meet any given load pattern. If you wanted to find the optimal mix you could treat it as a programming problem. This should also give you shadow prices (values) for the different resources.
    We now recognise that coal has significant external costs associated with it. Once these are added into the programming problem together with associated storage costs the solution to the optimising problem is (my guess) likely to be very different from the historical baseload and dispatchable problem.
    Thank you for making me think about this problem.
    David Murray

  5. I am no expert and I very likely missed something, but most systems currently that give baseload power lose much efficiency if not kept in steady state operation. With your spreadsheet for the reduced baseload scenario there are significant regions where much make-up power is required. Energy storage is always a lossy endevor as well. I see reference to “dispatchables”. Does this refer to storage of alternative energies? If so is not the devil in the details? Don’t get me wrong I am a big proponent of CSP.

  6. You’re exactly right about baseload power; that is the point of htis article.
    “Dispatchables” refer to any electici generation which can be ramped up and down quickly: CSP with thermal storage, natural gas turbines, hydro, and other generation when combined with electricity storage.

  7. Tom,
    I stumbled onto your blog via a link on the Solar Reserve web site. I’m a 27-year power industry veteran who now runs the power program of the European Climate Foundation, where we’re trying to fully de-carbonize the power sector by 2050 using predominantly renewable sources. Your emphasis on dispatchability is spot on, but you draw a false distinction when you pit baseload supply against dispatchable supply. You need to distinguish between baseload demand and baseload sources of supply. Both graphs in fact have exactly the same amount of baseload demand – that level of demand that we know must be met 24/7. Most baseload demand is met by dispatchable sources of supply (the exception being must-run sources like nuclear, run-of-river hydro, cogeneration plants, wind and PV – indeed, many of the low-carbon supplies we say we want, all of which as must-run resources are considered to serve baseload demand in a typical system load duration curve). The rapid response supply you depict in green in your curves does indeed come from dispatchable sources, but the most economic dispatchable sources are the same dispatchable sources that meet much of the baseload demand. Why? Because the cheapest and fastest response source of dispatchable supply is a large hydro or coal plant running at part load that can ramp up as needed on very short notice (most large hydro and coal plants operate at near baseload efficiency down to around 50% of rated load). Gas combined cycle plants can do this as well, but they’re not as flexible (usually because of the constraints imposed by the gas supply system) and they’re not as efficient as coal and large hydro operating at part load. Some dispatchable supply comes from plants that sit around waiting to be started up (mostly small, simple-cycle gas turbines), but they are nowhere near as fast in responding, they are very inefficient, and it is expensive to keep them idle for what is often 8500 hours a year or more. A system that greatly expanded its reliance on such peaking plants would be exceedingly inefficient in its use of gas and would put enormous strains on the gas production, transportation and storage system. The bottom line is that the dispatchable supply you are so blithely expanding and contracting must come in large part from plants that need to be operating at at least some minimally feasible level more or less all the time in order to be in a position to provide that dispatchable supply on demand. That doesn’t make them baseload plants, it makes them dispatchable plants that, because of their marginal cost of production, supply baseload demand some of the time and intermediate and even peak demand the rest of the time. If you make it infeasible for such plants to operate at a minimal level of output a minimum number of hours in the year, you will make the system untenably unstable because you won’t have enough sources that are responsive enough fast enough in all the right locations when you need them. You cannot talk about this topic usefully without an appreciation for the relative physical and economic operating characteristics of the various supply sources you are expecting to meet load at various points on your graphs. CSP, by the way, is also not particularly responsive if it’s been shut down – it would probably take the better part of an hour to start a CSP that’s in hot shutdown mode, and several hours for a CSP in cold shutdown. CSP’s value is indeed in its dispatchability, but to be truly the kind of responsive supply depicted in green in your graph, CSP would need to be operating at some level at the time that it was called upon. That doesn’t make it baseload supply, it makes it a dispatchable supply that can meet baseload demand for much of the time and provide responsive supply the rest of the time.

  8. Michael,
    Thanks for your detailed comment. I think you raise an important issue, although it’s one I chose not to address directly in the article. The distinction I am trying to draw is not between which types of generation are dispatchable and which are not; as you point out, most types of generation have some characteritics of both. CSP, for instance, is not dispatchable at all if it does not include thermal storage.
    So, as you say, we should not lump any sort of generation all into one category of “dispatchable” or “baseload,” but, as I expressed in the article, we should value a resource more the more quickly it is able to ramp up and down without drastic losses of efficiency.
    It’s not a question of a resource being “dispatchable” or not, it’s more of a continuum of how quickly a reasource can ramp up and down.


Please enter your comment!
Please enter your name here

This site uses Akismet to reduce spam. Learn how your comment data is processed.