This group brings together the best thinkers on energy and climate. Join us for smart, insightful posts and conversations about where the energy industry is and where it is going.


Levelized cost comparisons help explain value of various electric generation technologies

U.S. EIA: Today in Energy's picture
US Energy Information Administration

The U.S. Energy Information Administration (EIA) collects, analyzes, and disseminates independent and impartial energy information to promote sound policymaking, efficient markets, and public...

  • Member since 2018
  • 703 items added with 458,101 views
  • Jun 5, 2015

Levelized Cost Projections by Technology, 2020When power plants are built, several factors influence the choice of fuels and technologies that will ultimately generate electricity. Cost is one of the most difficult factors to compare, as technologies can have vastly different capital, fuel, maintenance, and financing costs, as well as different utilization rates and access to fuel resources. Two measures, the levelized cost of electricity (LCOE) and the levelized avoided cost of electricity (LACE), are widely used to make cost comparisons across technologies.

LCOE represents the per-kilowatthour cost of building and operating a generating plant over an assumed financial life and activity level (e.g., baseload, peaking, seasonal). Key inputs used to calculate LCOE include capital costs, fuel costs, fixed and variable operations and maintenance (O&M) costs, financing costs, and an assumed utilization rate for each plant type. These costs can vary by region and over time.

For example, technologies such as solar and wind have no fuel costs for generation and relatively small variable O&M costs, so their LCOE is mostly determined by capital costs and financing costs. Capital costs include things such as plant installation, labor, and grid-interconnection; financing costs are the costs of servicing the debt incurred during the lifetime of the plant. For generators that consume fuels such as coal and natural gas, both fuel costs and capital cost significantly affect LCOE.

LCOE values may also vary across regions because of differences in construction, fuel, and transmission costs, as well as differences in the quality of resource available for certain renewables such as solar and wind.

There are some attributes that LCOE does not capture, such as environmental considerations and grid operation constraints for maintaining system reliability. System reliability can be important, as some technologies are not dispatchable, meaning they cannot generate electricity on command at any time of day.

LACE, a different but related concept, represents the value to the electric grid (measured in per-kilowatt-hour terms) of adding generating capacity using a specific technology to the system. LACE reflects the cost that would be incurred to provide the same supply to the system if new capacity using that specific technology was not added. A technology is generally considered economically competitive when its LACE exceeds its LCOE.

Comparisons between the LCOEs of different technologies will not necessarily provide a good indicator of their economic competitiveness. This is clearly evident where new capacity using one technology competes against an existing plant using another technology. Because the cost of the existing plant has already been incurred, the new capacity addition will increase system cost unless its LCOE is lower than the operating cost (fuel and maintenance) of the existing plant. Many existing plants that were expensive to build but are cheap to operate, such as large coal and nuclear plants, can therefore be very economically competitive even though the LCOE for new plants of these types may be higher than the LCOE for other technologies.

Even when two new plants using different technologies are compared, LCOE may not account for differences in the services that each technology provides to the grid. For example, coal, nuclear, and natural gas combined-cycle plants all provide baseload services to the grid, and thus all have very similar LACE values, even where their LCOE values differ. In some regions, wind plants often provide higher output during the night when the demand for and the value of electricity is typically low. Wind plants also provide energy intermittently, and they cannot be ramped up and down by system operators to follow demand patterns. For these reasons, in a situation where wind has a similar LCOE to a combined-cycle plant, it typically provides less value to the system, as measured by LACE. Solar plants tend to produce most of their energy during the middle of the day, when the demand for and the value of electricity is higher, resulting in a higher LACE.

As part of the analysis that determines capacity additions in the Annual Energy Outlook 2015, EIA calculates LCOE and LACE for multiple technologies on an annual basis through 2040 in 22 regions of the country. These calculations are available in the full Levelized Cost and Levelized Avoided Cost of New Generation Technologies in the Annual Energy Outlook 2015 report.Levelized Cost Projections by Technology, 2040

U.S. EIA: Today in Energy's picture
Thank U.S. EIA: for the Post!
Energy Central contributors share their experience and insights for the benefit of other Members (like you). Please show them your appreciation by leaving a comment, 'liking' this post, or following this Member.
More posts from this member
Spell checking: Press the CTRL or COMMAND key then click on the underlined misspelled word.
Nathan Wilson's picture
Nathan Wilson on Jun 6, 2015

So the US government EIA has updated their forecast for electricity cost by source, and finds that not only do solar and wind require subsidies today (in most places), but they will continue to need them through 2040 (clearly off-shore wind, but also on-shore, which is under half the cost, and will contine to have an LCOE above the avoided cost in 2040.)

This assessment stands in stark contrast to the proclaimations of advocates who say they are competitive today and dropping rapidly.  So clean energy subsidies/incentives are still needed.  In fact, the EIA projects that on-average in 2040, nuclear will need a smaller subsidy than solar, hydro, biomass, or off-shore wind (see tablular data here).

Joe Deely's picture
Joe Deely on Jun 6, 2015


REALLY surprised that you fell for this crap. If you are willing to do some reading – you can see why it is crap. 

Historically these EIA reports have not been remotely close in predicting solar costs – even when only going 5 years out. See here and here for two examples from 2010 and 2011 Energy Outlooks. These both show predictions for 2016. Average cost per megawatt hour decreased $186 in one year! Now they are trying to predict solar out 25 years. 

The problem is that their simple model cannot handle technologies that are rapidly changing.  To see this you need to dig into the model assumptions.

There is an assumption file for each Outlook year. Here are the assumptions for the 2015 Outlook. Note: for some reason it appears they have not been updated from 2014 yet. You can go back to previous years and find the assumptions for those years as well.

A big part of the LCOE calculation is based on actual capital costs for that technology in the previous year. Makes sense for coal, gas and nuclear – not so much so for solar. Click on the Electricity Market module to see the capital cost numbers for solar.Y ou can see in Table 8.2 the cost for a 150MW solar plant built in 2013 was $3,564/kW. Seems a little high to me, but more imporantly this is down from $4,755/kW in the 2011 Assumptions. 

I would assume that the 2015 LCOE is using capital cost data from 2014 , but since the assumption file is still labelled 2014 –  that may be an incorrect assumption. According to the most recent SEIA report , utility scale installation costs were down another 20% in 2014 from 2013.

Obviously for 2040, the EIA does not have a recent year to look at capital costs. So they have to rely on modeling capital costs and learning curves. 

  • “Technologies and their components are represented in the NEMS model at various stages of maturity. EIA classifies technologies into three such stages: mature, evolutionary, and revolutionary. The technology classification determines the rate of cost reduction that can be achieved through the learning function. Generally, overnight costs for technologies and associated components decline at a specified rate based on a doubling of new capacity. The cost decline is fastest for revolutionary technologies and slower for evolutionary and mature technologies. “

Here and here are two places to go for further information on this part of the model. In particular the second location gives a deeper dive.

In reading this material, I was surprised to see that the EIA only considers(Table3) PV solar to be an  evolutionary technology. It seems like the cost decline of solar has been revolutionary to me.  

For the last six years the EIA has had to dramatically change their LCOE assumptions for solar each year because of rapidly falling capital costs. This can easily be seen the assumption file for each Outlook year. However, magically between now and 2040 those capital costs are going to stay flat. I don’t think so. 


Clayton Handleman's picture
Clayton Handleman on Jun 6, 2015

Nathan, you seem to have fallen to that ‘confirmation bias’ that you often warn against.  Joe is correct, EIA has been so far off for so many years when it comes to renewables projections that I believe the system to be heavily influenced by FF and or Nuclear interests. 

While I think you have a number of helpful ideas, you seem so eager to push your agenda that you will grasp at anything that will support it.  EIA is good at gathering data.  They are a total fail when it comes to making projections for renewable energy.  I am disappointed to see you relying on their projections.  Ouch.

Bruce McFarling's picture
Bruce McFarling on Jun 7, 2015

This assessment stands in stark contrast to the proclamations of advocates who say they are competitive today and dropping rapidly.”

Actually, for wind it is more confirmation of the “competitive today” than stark contrast  … as revealed when after you say “not only do solar and wind require subsidies today”, you then immediate qualify it with, “(in most places)”. Given the size of the economic resource, “a few places” where wind is cost-competitive is quite a lot of electricity generation potential.

And of course, these are EIA projections, so these are not feasible costs under a best case, but projected costs under current policy. A regionalized system of transmission planning would substantially increase the market potential for the better wind resources.

And of course, this is measured against underpriced fossil fuel electricity. The EPA benchmarks are 890lb/MWh for advanced NGCC, which is about 0.4tonne, so at $50 per tonne CO2 ought to be charged an additional ~$20/MWh for all natural gas at the EIA multiple of 1.21lb/kWh, and at $100/tonne ought to be charged an additional ~$40. At the optimistic $50/tonne for the external cost of the NGCC, the $66-$81/MWh LCOE for advanced NGCC would rise to $86-$101 in 2020, and wind, solar and run-of-river hydro have a substantially wider selection of competitive resources, and advanced nuclear and possibly sustainable biomass come into the frame (“possibly” because, while unsustainable biomass loses its cost advantage over sustainable biomass as carbon prices rise, EIA cost estimates are largely for unsustainable biomass, which would see LCOE rise with carbon pricing).

As noted already in the comments, one cannot credibly use EIA 2040 projections to contradict the claim that the cost of solar is dropping rapidly, given the EIA’s track record as a persistantly biased over-predictor of future costs of solar.

Keith Pickering's picture
Keith Pickering on Jun 7, 2015

Dear EIA,

I do understand why EIA (and others) fail to include plant lifetime in LCOE analysis: it’s because you’re taking an investor’s viewpoint. In other words, if you were a utility, what’s the bottom line for building new generation technology? And if you’re a utility, LCOE is a good metric to help answer that question, because the only lifetime you’re considering is the financial lifetime, i.e., the lifetime of the loan, not of the powerplant. Which in your analysis is always 30 years.

But not everyone is a utility, and the investor’s bottom line isn’t necessarily what’s good for society as a whole. As a government agency, you should at least consider taking a broader view, and including the effect of plant lifetime in your LCOE analysis. Perhaps we should call it something else, so people understand the difference, like LLCOE (Lifetime Levelized Cost of Electricity). A constant 30-year assumptions for all technologies is far too simplistic, and LLCOE would correct that. The 30-year constant unfairly penalizes long-lifetime technologies (like nuclear, and especially hydro) while unfairly benefiting short-lifetime technologies, like wind.

One reason wind looks really good from an LCOE standpoint is because you’re assuming a 30 year loan lifetime for a device that won’t last that long (unless it’s in a rather non-windy location).Frankly, its rather doubtful that any banker who does his homework would give 30-year terms on a loan for a wind turbine. LCOE should reflect that, but it doesn’t.

Another part of LCOE where you fall down is systems costs. You guys do make some attempt there by including transmission investment (necessary, considering that the best renewable sites are far from cities). But transmission alone isn’t the whole story, and others are beating the pants off you in this regard. In particular, OECD recognizes four types of systems costs often neglected in LCOE compuations, and there’s no reason you shouldn’t either: backup, balancing, grid connection (roughly similar to the transmission investment in your calculations) and grid reinforcement.

So the way I see it, your LCOE computations are about as good, and about as bad, as anyone’s. But they sure could be better, and it wouldn’t be that hard to fix.

Your friend,

Keith Pickering

Nathan Wilson's picture
Nathan Wilson on Jun 7, 2015

Hey, sorry if it sounded like I was endorsing the EIA projections for 2040, that was not my intention.  I don’t have faith in anyone’s predictions that far out.  But I did want to call attention to the fact that all of the professional predictions for renewable cost don’t show the aggressive price drops that are the only one we see discussed in the green articles. 

Clayton Handleman's picture
Clayton Handleman on Jun 7, 2015

“all of the professional predictions” 

Honestly, existing studies are more than sufficient to get a pretty good handle on this.  And some, such as SunShot below, have stood the test of time, i.e. their projections are turning out to be pretty much on target several years after being made.  It may be hard to predict exact rates but easy to see that there is a good deal of headroom for significant cost reduction for the forseable future. 

– Are you rejecting Sunshot?  Which professional projections are you referring to and which ‘Unprofessional’ projections are you rejecting.

When I first ran across the SunShot projections I was somewhat skeptical as well.  However, when China collapsed the prices of PV and SunPower and FIRST were able to stay in the game I realized we really were going to do it.  With module prices much lower, they are not the only cost driver.  The bad news is that the other cost drivers are more difficult to drive down.  The good news is that there is a fair amount of headroom.  One need only look to Germany where, with higher labor prices, they are getting much lower installed pricing than we have.  There are a few areas for improvement:

1) Balance of systems – get the inverter, racking and wiring costs down and you bring down the system cost.

  – Inverter costs: Like modules, technology and volume improve the cost.

  – Racking: Experience curve, solar ready building codes and increased module efficiency drive down costs.

  – Soft costs: This is where Germany has done a spectacular job in the residential sector.  They have gotten their codes and standards in shape.  More importantly they have streamlined their system approval process.  In many parts of the US the wiring inspectors aren’t even up to speed on the solar part of the NEC.  This despite well over a decade since it became a complete and mature document. 

2) Modules

  – Module cost $/W:  Volume is high enough that it will be harder to get the cumulative doublings but there is still headroom.  For each cumulative doubling, the cost comes down 17%.  That has held true for over 3 decades

  – Module cost efficiency:  Modules are still getting more efficient.  This is a triple win, you get the improved module cost on a $/W basis AND you get the reduction in BOS AND you get the soft costs amortized across a larger system so they cost less on a $/W basis.  Consider that if you go from a 15% (typical in today’s modules) module to a Sunpower 21% module that you get a 40% improvement in space efficiency.

   For some time SunPower was the only game in town for high efficiency modules.  However

   FIRST Solar has been ramping up the efficiency curve rapidly with the potential to, yet again, disrupt the PV industry.

  In an effort that is less publicized than his battery factory, Elon Musk is building a solar gigafactory.  The plan / expectation is that they will be producing PV modules at near SunPower efficiencies with lower $/W costs.

Voltage Increase:  The industry is working towards a 1000V module standard.  There are a variety of esoteric reasons that this will improve overall system efficiency and reduce costs.  One is that it allows the inverters to squeeze more efficiency out of the already near miraculous SIC semiconductors by making their inherent voltage drop a smaller fraction of the operating voltage.  The other is that losses go as the square of the current.  By nearly doubling the voltage you half the power.  This leads to cutting wiring losses to roughly 1/4 of what they had been.  Designers win by lower losses = more power delivered or less wire (copper is expensive so this is a nice cost reduction).  Or some optimal combination of both.

Bottom line – solar cost reductions are not played out.

I am less versed in wind but can speak to some aspects.  The experience curve for windpower is 14% per cumulative doubling all other things being equal.  As I understand it that is for the cost of the machines before taking into account other aspects.  Two other areas that are important are Capacity Factor (CF) and maintenance costs. 

  – CF: Average CF is below 40% and as I have mentioned frequently in comments sections, with moderate transmission build-outs (some of which are in process as we speak – TX and KS) wind can be sited in 50% CF sites.  My preliminary calculations show that the 25% improvement in CF would pay for large High Voltage DC conduits to the high use areas such as the Eastern costal cities.*  Clean Line Energy certainly thinks thinks that there is a value proposition here.  They are developing multiple HVDC lines and my understanding is that they feel the economics works without subsidies.

  – Maintanence Costs:  In the annual wind reports with which you are no doubt familiar, maintanence costs have been dropping.  Basically it would appear that the industry has cracked this nut.  There were problems a while back but they appear to be solved, lower down time, and less cost.  Some of the backstory is interesting for tech geeks.  Things such as increased quality of lubricants, premptive diagnosis and repair of damaged components, higher quality components in critical wear areas, managing blade tip wear etc.  The industry is large enough that companies can dedicate full time engineering efforts to solving these problems and they are succeeding. 

*There are some who repeat ad nauseam the tired line that there will be terrible visual disruption due to power lines being strung across the US.  First, unlike the highway system which we managed to built, there is ample precident for power lines.  Second +/- 800kV HVDC has very high power density, more than an order of magnitude higher than the 345kV which is the most common go-to for bulk power transmission.  And farmers get paid for siting it on their property so it will be interesting to see how much resistance there really is within the envelope of existing law.

Bruce McFarling's picture
Bruce McFarling on Jun 8, 2015

On wind, in addition to the factors that you mentioned …

Note that if you aggregate wind from wind farms spread more widely across a single wind resource, and from distinct wind resources that lie to the east and west of each other, the total capacity that can be integrated is increased, because the wind does not blow strongly in all of the wind resources at the same time.

If the Nameplate Capacity is N, the Maximum Yield is M, and the Average Yield is A, the argument that the greatest share of energy you can get from that variable renewable is (A/N) is mis-stated … it is, rather, (A/M), the ratio of average yield to maximum yield. If average capacity factor is 45% and maximum yield is 80%, then the rule of thumb integration maximum is not 45%, its 56%.

If we look at the duration curve, the losses to over-capacity investment increase roughly in proportion to the square of the overinvestment, since the greater the overinvestment, both the greater the curtailed harvest when overproducing and the more hours of the year that is being over-produced. Which, by the same token, implies that a modest degree of over-investment has a relatively small cost. If a system curtails between 0% and 10% of nameplate capacity about 30% of the time, that is on the order of a 1.5% of total nameplate capacity hours, which with a CF of 45 is about 3.3% of total available energy. Raise the ceiling from the rule of thumb to 110% of load, multiply by 1.25 to account for a (M/N) of 80%, and the rule of thumb rises to 60%. 

If the pricing system is not set up to sabotage the returns to windpower with modest amounts of over-capacity investment, a modest amount of total available energy supply in excess of load is a reasonable cost-optimizing strategy.

And while the dueling studies approach to discussion of windpower leads to common assumptions that available capacity factors are fixed, the engineering reality is that capacity factor is in part a design decision of how large a blade cross section to connect to how large a generator … similar to the way that higher CF, so lower cost per kWh, or higher nameplate capacity and so greater revenue from selling into peak load and provision of grid services is a design choice for reservoir hydro. The same 3MW generator in the same location with different length blades and different nacelle heights will have different capacity factors.

If we charged a carbon price for use of the scarce natural resource of natural ability to sequester atmospheric CO2, the 2020 LCOE vs LACE comparison above, without substantial carbon pricing, demonstrates that there would be substantial design leeway to increase the CF for a substantial range of possible windpower sites.

Get Published - Build a Following

The Energy Central Power Industry Network is based on one core idea - power industry professionals helping each other and advancing the industry by sharing and learning from each other.

If you have an experience or insight to share or have learned something from a conference or seminar, your peers and colleagues on Energy Central want to hear about it. It's also easy to share a link to an article you've liked or an industry resource that you think would be helpful.

                 Learn more about posting on Energy Central »