Solar PV capacity factors in the US – the EIA data

A post I wrote a little over two years ago concluded that solar PV capacity factors in the US ranged between 13% and 19% with an average of around 16%. Recently, however, the US Energy Information Agency published a table showing an average capacity factor of around 28% for utility-sized PV plants in the US in 2015. This post looks into the reasons for this large difference and also addresses the question of whether the EIA estimates can be used to predict future US solar PV output.

It concludes that the EIA estimate for utility-scale plants is probably overstated by several percent and that the overall capacity factor is further overstated because the EIA does not take smaller (mostly rooftop) arrays, which have a significantly lower capacity factor, into consideration. Assuming that future PV plant construction in the US is evenly split between utility-scale and “distributed” installations, and that large plant construction takes place all over the country and not just in California, an overall capacity factor of around 20% would be a reasonable assumption for planning purposes.

Data Sources:

Unless otherwise specified all the data used in this post are from the zipped EIA tables accessible through this link . The data are preliminary and the EIA cautions that they should not be summed, but having no other data to work with I summed them anyway.

All capacity factor estimates are generation-weighted. Note also that the EIA data apply only to “utility-sized” (1MW or greater) solar PV installations, so a large number of rooftop solar and other “distributed” solar installations are not included.

The EIA 2015 data include generation and capacity data for approximately 1,250 solar PV plants, but a number of them do not have complete data for the year and some of those that do give implausible capacity factors. After weeding these out I was left with 942 plants that could be considered to provide valid data. Basic statistics for these plants in 2015 are:

  • Installed capacity: 8,290 MWp
  • Annual generation: 18,291 GWh
  • Capacity factor: 25.2%

Data Review:

The 25.2% capacity factor listed above is about 3% lower than the ~28% calculated by the EIA from (one assumes) the same data set. As shown in Figure 1 this is an across-the-board effect, with the difference remaining reasonably constant at between two and four percent in all months:

Figure 1: Comparison of EIA monthly capacity factors and RA monthly capacity factors estimated from EIA individual plant data, 2015

What accounts for this difference? I went through the supporting EIA pdf but was unable to find any combination of capacity and generation data that allowed me to replicate the EIA capacity factor estimates. Another possibility is that my results are skewed by capacity additions during the year (I assumed the same capacity in each month), but this would not explain the constant difference between the plots. Besides, December 2015 generation was only 8% higher than January 2015 generation, indicating that there were no major capacity additions during the year.

However, the EIA data show a clear dependence of capacity factor on plant capacity, with plants larger than 50MW having higher and more consistent capacity factors, as shown in Figure 2. It’s possible that the EIA might have given more weight to these larger plants when calculating averages, but that’s speculation on my part:

Figure 2: XY plot of capacity factor versus installed capacity for 942 plants in the EIA data base, 2015

The average capacity factor of approximately 18% at 1MW is also broadly comparable to the 16% estimate in my previous post, which was based dominantly on plants less than 1MW in size. More about this later.

Using the EIA estimates as an average for the entire US is also invalid from the geographic standpoint. A disproportionate amount of megawatts of capacity and gigawatt-hours of generation come from plants located in California and in adjacent parts of Arizona, all of them in a comparatively small desert area where PV capacity factors are the highest in the country. (All of the plants with capacities of over 100MW shown in Figure 2 are in California or Arizona.) Table 1 summarizes the EIA data for each state with ten or more PV plants, ranked by capacity factor. The 12 states with fewer than ten plants are lumped together in the “Other” category:

Another approach is to compare the “Sun Belt” states (Arizona, California, Florida, Hawaii, New Mexico, Nevada and Texas) with the “Others” (Table 2). The solar-favorable areas of the Sun Belt states probably make up less than 10% of the total area of the US yet they generate over 80% of the country’s solar PV electricity.

Figure 3 is a map of the US lower 48 summarizing the Table 1 data. Capacity factors decrease by up to 12% going from southwest to northeast. Coverage is quite good except in the Plains States, Montana, Wyoming and Idaho, where there are no data:

Figure 3: US state map summarizing the Table 1 results


Earlier I noted that the average solar PV capacity factor of approximately 18% at 1MW is also broadly comparable to the 16% estimate in my previous post, which was based dominantly on plants less than one MW in size. Figure 4 superimposes the capacity factors from the previous post on the EIA data out to 10MW. Visually the two data sets appear to line up. In combination they suggest a gradual decrease in capacity factor from about 21% at 5MW to about 16% in the low kilowatts range.

Figure 4: XY plot of 2015 EIA individual plant data with individual plant data from RA (Roger Andrews) 2014 post superimposed

The most likely explanation for this decrease is that utility-scale PV plants are sited in open, sunny areas with the panels pointed in the optimum direction and often with axial tracking capability. Residential and business-scale arrays, however, usually consist of fixed roof-mounted panels that point the way the roof points, which often isn’t the optimum direction. Improvements in PV panel efficiency may also have contributed to an overall increase in capacity factors over the last few years but it’s not possible to quantify the contribution from the available data.

The remaining question is what the overall capacity factor for new solar installations in the US is likely to be. There are of course a large number of variables that have to be taken into account, with the most important being:

  1. What proportion of the new capacity will be utility-scale and what proportion residential-business scale?

In 2015 utility-scale solar made up 58% of total installed solar PV capacity and residential-business solar the remaining 42%, according to the EIA data. I have assumed a 50-50 split in the future.

   2. Where will the new capacity be built?

Up to now it has been concentrated in the sunny Desert Southwest, but capacity additions in this area will eventually become limited by the size of the Southern California/Arizona market, or by technical obstacles such as the California Duck Curve, or possibly by environmental constraints. For the US to “go solar” in a big way more utility-scale plants must be built close to major consumption centers in the eastern and central parts of the country and many households and businesses in these areas must be induced to install solar panels on the roof. I have therefore assumed that half of future US solar PV capacity will be installed in the Sun Belt states (developers will still have an incentive to go after the better resources) and that the other half will be installed outside the Sun Belt states, mostly close to centers of consumption in the Northeast US.

These two assumptions yield four sources of new solar capacity, each accounting for 25% of total capacity:

Utility-scale capacity in the Sun Belt with an average capacity factor of 27.5% (Table 2)

Utility-scale capacity outside the Sun Belt with an average capacity factor of 18.4% (Table 2)

Residential-business scale capacity in the Sun Belt with an average capacity factor of 17.5% (estimated from a segregation of the Figure 4 RA data)

Residential-business scale capacity outside the Sun Belt with an average capacity factor of 15.1% (from the same segregation)

These four sources have an overall capacity factor of (27.5+18.4+17.5+15.1)/4 = 19.6%. Twenty percent would therefore be a good round number for planning purposes, always assuming that US solar subsidies are maintained at the levels necessary to support further PV expansion.

This entry was posted in Energy and tagged , , , , . Bookmark the permalink.

40 Responses to Solar PV capacity factors in the US – the EIA data

  1. Willem Post says:


    The VT-DPS has a website that lists the mostly production, MWh, of all larger PV systems (mostly field-mounted) in Vermont, for the past 6 years.

    My spreadsheets show the average CF of those systems is about 14.6%

  2. Euan Mearns says:

    Figure 1 is very strange Roger, and at face value suggests that the anomaly is not down to capacity additions in current year. But the EIA could perhaps still be making an invalid adjustment for new capacity additions?

    I’m surprised to see such a large difference between Texas and California. Does cloud cover data support this?

    And I’d finally note that capacity factor in the UK is:

    Table 1 Summary of generation statistics from National Grid. Installed capacity from REF. The National Grid generation data combined with the REF installed capacity data yields a solar PV load factor of 10.3% for the whole year. This compares with 10.8% published by DECC and 11.8% determined from REF data (see A Note on UK Renewable Load Factors) and with 10.1% estimated by Roger Andrews.

    We Brits can do our sums right 🙂

    • Alex says:

      From an analysis of national grid data or an up coming report:

      Solar (and wind) CFs by year:

      But worse for solar – though not wind – in the UK, is the temperature correlation (data added from HadCRUT central England:

    • Thinkstoomuch says:


      Remember Texas is a huge state. Something like 800+ miles east west and 600 miles north south. With wildly divergent weather patterns. Desert (El Paso) like to coastal humid and rain (Houston). Just a couple of highlights. 10 data points is a sparse data set.

      It actually gets worse if you drive through. Pay attention to the fuel gauge and where you are going unless you like walking a long way when you run out. 180 miles is not uncommon for next fuel.

      Look at the first picture, at the top of the post, of the kwh/day variation. It varies from 8.5 to around 7. CA another very large state(if not quite as big) has much the same distribution of kwh/day. But people live close, relatively, to where the sun shines. Not so much in Texas

      A couple of graphics from the EIA as to where the solar plants are located.



      Texas has very few plants in relation to its size and fewer are in the “prime locations”. California has for the most part concentrated the utility scale solar in the desert areas that are at less than 100 miles from the population centers.

      Not knocking what Roger did(it is great and incredibly better than I could do) and is one of the things that generally makes me think in circles when looking at statistics and my inability to deal with this stuff effectively. Especially as to 50 different states worth of stuff then comparing it to other places.

      Hopefully this post makes some kind of sense,

      • Roger Andrews says:


        Yes, Texas is a big state. I’ve driven across it a few times and it takes a while.

        Of the twelve Texas plants in the EIA data base two are located in prime solar areas (El Paso and Presidio). They have an average capacity factor of 29%. The remaining ten are located in non-prime solar areas in and around San Antonio and Austin, and nine of them have an average capacity factor of 20.7%. The remaining system is rooftop-mounted on top of an IKEA store outside Austin, which explains the lower capacity factor (16.4%).

      • jim brough says:

        I looked at which showed the distribution of solar energy generation in California. It covers a great spread of latitude which ultimately determines the amount of solar energy reaching the collectors.
        I think that the data should show that the capacity factor will be lowest at the Northern end of California. Latitude is important but not the only factor. Can someone provide that ? At age 83 I find it difficult to dig into the data.
        Australia has a sunny aspect but the establishment of reliable solar electricity supply to the capital, Canberra, is not good when the city experiences regular fog.

        • Roger Andrews says:

          Jim: The inset at the top of the post tells the story. Southern California gets 8.5 kWh/day of solar radiation and the extreme north less than 7kWh/day. So we would expect capacity factors in Southern California to be 4-5% higher than capacity factors in Northern California, all other things being equal.

          It’s also refreshing to come across someone who’s even older than I am.

    • Roger Andrews says:

      Euan: One possibility that I didn’t mention but should have is that EIA estimates capacity factors using MW(ac) rather than MW(p). MW(ac) will give slightly higher capacity factors because it allows for losses in the inverter. The EIA does, however, give all of its capacity numbers in MW(p).

  3. Alex says:

    One way to boost capacity factors is to install a smaller inverter. This loses a small amount of PV at peak time (when wholesale prices are negligible). It also reduces the capacity and hence boosts the capacity factor.

    It could be that subsidy bands push this down? In the UK for example, the high subsidy capped at 4KW. Could operators install 4.5KW of panels, de rated to 4KW, and a 4KW inverter?

    • It doesn't add up... says:

      What happens to the extra 500W?

      An acquaintance had her house burn down because their new installation didn’t have the inverter correctly connected. It’s taken them 5 years to get it rebuilt, with extensive arguments with insurers, the installer, etc.

      • mbe11 says:

        It is called having a fuse on the supply so it does not exceed the capacity of the inverter when you draw current.

        • Thinkstoomuch says:


          A nit pick but that is not how a fuse(or circuit breaker) works. Fuse basically lets electrons through or doesn’t. Zero power or full power.

          Power equals Voltage times Amperage.

          My little panels don’t care what is connected and what the load is drawing. Nothing connected no current, infinite resistance (blown fuse scenario), no power.

          Current or voltage limiting requires a lot more circuitry than a simple fuse. Though a circuit breaker is probably incorporated.

          For Alex’s Question:

          US Federal subsidies (generally they are cost based tax credits) in the US don’t seem to reward for generation anymore (I think). The individual states may, or contracts may. But I don’t think it is as much as a player as in Spain based on prior Energy Matters posts and analysis.

          Much like that map at the top just better solar environments. Compare that to just about anywhere in Europe and see what it looks like. Of course make sure of the scale and what is actually measured.

          Have fun,

      • mbe11 says:

        You install a fuse on the inverter to limit current draw.

        • Alex, and mbe11. Under sizing of inverters is recommended by Enphse on their systems to improving the Wh/dollar performance. By mating a 250 Watt panel with their M-215, which will only put out about 225 Watts, loses at the peak on clear cold days are more than made up on the shoulders, and the fact that in most installations there won’t be that many days where it is 25º C or less on the roof at noon. As for fuses, I would not wan’t an inverter that was not designed with the ability to handle more than its rated input without damage. It may not be able to convert all of the DC input available, but it must not cause damage.

          • Thinkstoomuch says:

            That is an interesting datum. Thank you.

            What is the cost difference of the 225 watt model with the 250 watt model? Or a 280 watt model?

            Also when I looked at current panel spec sheets most seemed to be rated for around 45 degree C. Not sure how that would impact the cost/benefit.

            For example the Astroenergy 260 watt peak panel:

            Normal operating cell temperature (NOCT) 43±2°C


            There is lot of neat stuff in spec sheets for someone who is sufficiently anal like myself. Like at STC it is only a 240 watt panel.

            Thanks in advance,

      • Alex says:

        Dangerous stuff, this solar power!

        I’m not sure how the “fuse” works – as clearly that 500W has to be consumed somewhere.

        I have a 5.4KW array and a 5KW inverter – though I calculate my capacity figures based on 5.4KW. I sometimes see some “shaving” on very sunny, cool days, but it makes little difference to output.

        If the subsidy cut-off was 5KW, then I’d be tempted to call it a 5KW system .

        • OpenSourceElectricity says:

          Limiting the Inverter to 70% of the panel capacity is standard for rooftop solar in germany. The losses have been calculated as 3% of the electricity produced by the panels, which is more than compensated by the lower costs of the inverter. The inverters can handle this.

  4. Thinkstoomuch says:

    Thanks for another very informative and thought provoking post!


  5. Peter Lang says:


    Very interesting, thank you. Are you able to separate out the generation from plants hat have solar tracking? It would be interesting to see what the CF of just the fixed array commercial plants. I suggest exclude all plants where the generation is not metered and reported – therefore,exclude residential and un-metered smaller plants. I realise this is not average, but I am seeking to understand the upper bound (with current technology) for fixed array commercial plants.

    • Roger Andrews says:

      Peter: EIA gives information for 12 solar PV plants with capacities of over 100MWp. Eight of them have single-axis tracking panels and four have fixed panels. It’s interesting to note that the three plants with the highest installed capacity (Desert Sunlight, Agua Caliente and Topaz – see Figure 2) all have fixed panels. This probably explains why their capacity factors are low relative to the smaller utility scale plants.

      • Peter Lang says:


        Thank you. That helps. So it would seem the large fixed arrays are probably tending to dominate, but there is sufficient from the tracking plants to make the CF higher than I would have expected. Are the solar towers included? Do some of them have energy storage and using gas to help them. It would be interesting to separate the different types to allow the capacity factors to be compared.

        I agree with your comment to Dave Rutledge that the denominator for calculating capacity factor should be the AC capacity the plant can supply to the grid (net output as is conventional practice for all other plant types).

        • Peter Lang says:

          I should have said “sorry, I don’t have time to read the EIA report at the moment.” But perhaps other followers of Energy matters might be interested in these issues too.

        • Peter: The data are for PV installations only and don’t include CSP “solar tower” plants. Despite their claimed advantages CSP plants in fact have a lower overall capacity factor than PV plants according to the EIA table. I got similar results for the CSP and PV plants in Spain:

          • Peter Lang says:

            Thank you.

          • Thinkstoomuch says:

            Peter and Roger,

            I waited to post as I thought others would have more information.

            There are only three Solar Tower Plants in the US according to the EIA. A two tower 5 MW plant in the Mojave, Ivanpah (three towers) and Crescent Dunes in Tonapah, Nevada.

            I have been tracking those and a couple of others in a spread sheet. Comparison of CF’s graph.


            SS North PVSP is the Silver State North PV Plant just across the border from Ivanpah. A good check for it. A fixed tilt 52 MW summer capacity plant.

            Solana is right in the heart of the “good” spot in Arizona so not exactly a fair comparison. But not far off either.

            I think their are 22 or so CSP plants in total(includes the tower plants).

            Conceptually I have a serious soft spot for Solana. Unfortunately it is not economic. Which is why I was surprised that Australia is pursuing the towers, judging by recent headlines I am seeing.

            Have fun,

  6. Dave Rutledge says:

    Hi Roger,

    Excellent post. Some of the EIA capacity factors seem extremely high to me, given that they are for the entire United States. They list 34.8% for May 2016 for solar PV and 40.3% for wind in March 2016. I like Alex’s theory that the capacity factor is calculated from the inverter capacity rating, which may be conservative. My experience is that the AC output never gets anywhere near the capacity of the panels themselves.


    • Roger Andrews says:

      Thanks Dave: I agree that it would be better to express PV capacity as what comes out of the back side of the inverter rather than what goes in the front. This is, after all, what we do with thermal plants, where capacity is calculated based on the amount of electricity coming out of the generators rather than the quantity of heat going in. .

  7. climanrecon says:

    I have a very small solar panel at home, not on a roof, it can be moved around, for maintaining the charge of a car battery. I’ve noticed a very strong angular effect, the output current drops away very strongly when the panel axis moves away from the direction of the sun, one day I’ll measure it quantitatively, but if I had to guess I’d say the effect was much stronger than just the cosine of the angle.

    Capacity factor calculations may assume a cosine angular dependence, which may not be a good approximation.

    • Alex says:

      The other thing I notice is a 90% drop off when a thick cloud gets in the way.

      I quite often hear that “modern panels work on daylight, not sunlight”. That’s bollocks.

      • Greg Kaan says:

        Both true

        My issue with tracking panels is that although they increase capacity factor, the cut in and fall off rates at sun rise/down are greatly increased. This makes ramping of other generators to match a much greater task for the grid operators.

        As always, storage would be the solution if the capacity existed.

      • My solar panels begin to twitch at dawn before the sun even appears over the horizon. They continue to generate when a cloud obscures the sun, with the amount of generation dependent on the “blackness” and extent of the cloud. I also get the “silver lining” effect, where sunlight reflected off fluffy white clouds increases radiation over what it would have been in a clear blue sky. The spikes on September 12 on this solar radiation graph from our local weather station are an example.

        So to put the issue in perspective, solar panels do indeed work on daylight, but the daylight is provided by the sun and modified by clouds

  8. Roger Andrews says:

    A brief update: I mentioned in the post that future solar development in California might be constrained by environmental factors. There now seems little doubt that it will be. As will shortly be reported in Blowout Week 142, the US government has banned further solar development on 9 million acres of environmentally-sensitive public land in California, leaving the solar industry with only about half a million acres to play with. Unsurprisingly, the solar industry is not amused.

  9. sod says:

    This article is offering another explanation for the pretty high CF numbers in the USA:

    “There are a number of interesting changes mentioned in the report on solar energy. One is that the price of photovoltaic panels has dropped so much that it’s changing the way the plants are set up. We’re seeing more installations where the total direct current output can exceed the installation’s capacity to convert it to alternating current, which is needed before the electricity can be put on the grid. In other words, it now makes economic sense to buy more panels than are strictly needed, just to make sure your DC-to-AC hardware is kept at full capacity when the generating conditions aren’t ideal.

    Another thing that’s changing is that solar tracking devices are becoming more common—typically, these tilt the panels to follow the Sun as it moves from east to west over the course of the day. While this adds to the hardware cost, it has less influence on the price than other factors (including the cost of approval and permitting), while allowing the installation to produce more power. It also lets solar help service more of the period of peak load on the electric grid.

    Partially as a result of this, the average capacity factor (the ratio of the actual generation compared to the facility’s potential) of utility-scale solar is going up. While the overall mean capacity factor for the projects the DOE looked at was 25.7 percent, it had gone up to nearly 27 percent in more recent projects. This increase is coming despite the fact that more plants are being built outside of the Southwest, and thus in locations where the solar resources aren’t as great. (One project had a capacity factor of over 35 percent.)”

    A system that has a lower output level than it could have will have a higher CF. with the panels getting cheaper, this might make economic sense.

  10. Leo Smith says:


    The ratio of average delivered power, expressed as a percent, to whatever nameplate figure the manufacturer cares to place on the back of the equipment as its ‘rated maximum’.

    If you down rate the kit, and arrange for excess output to be discarded, you can get a much higher capacity factor, albeit at less overall delivered energy….

    In the case of PV its not really a very useful measure of anything. What is more important is watts per unit land area, or unit cost, and the peak to mean ratio which ought to be the capacity factor if that wasn’t so easy to manipulate

  11. Leo Smith says:

    …further to the above, as far as I know, if you fail to draw current from a solar panel array, nothing untoward happens. It has been mentioned that using an inverter that is unable to handle the peaks will give an overall better ‘capacity factor’ and, depending on the relative costs of panels and inverters, may actually prove to be better ‘watts per $’.

    Inverters will be innately designed to cope with variable input voltage, and current, and to drive at constant grid voltage but variable power.

    Its not a question of having a fuse and not delivering power at all. Its just (probably) software to map what’s available into what’s pushed out to the grid, with some smartness in terms of how hot the electronics is getting etc etc.

  12. Thinkstoomuch says:

    Thank you all for the thought fodder. Especially Roger for putting the original post together and making me feel better about not being able to figure out the EIA CF numbers!

    A couple of reports from August 2016 that lend some credence to the inverter sizing, tracking, module types and such. I am still digesting them. As well as some other stuff they sort of cover.

    Both have the report, presentation(36 and 37 pages) and data files. Second is focused on smaller systems especially residential.

    They seem fairly comprehensive for those interested. And they do mention that inverter loading is 1.31 in 2015 and 1.2 in 2010 for utility scale.

    Have fun all and Thanks again,

Leave a Reply