Estimating Global Solar PV Load Factors

Guest post by Roger Andrews:

In the recent Efficiency of Solar Photovoltaics post Euan Mearns presented some solar PV load factors calculated from BP 2012 installed capacity and generation data that made no sense, such as 7% for the US and 30% for Spain. So I did some work to find out whether these numbers were isolated instances. I calculated more load factors using 2012 data from other sources, such as Wikipedia, Observ’ER and Photovoltaic Barometer, added a few more countries and plotted the results against latitude. They are all over the map (Figure 1):

[Image – World’s northernmost solar PV system – Kotzebue, Alaska. Latitude 66.9N, load factor 8.9%]

Figure 1. Load factors vs. latitude calculated from national solar PV statistics, 2012.

And the reason they are all over the map is that different data sources give conflicting estimates for 2012 solar PV generation and installed capacity. Here are some examples:

  • Chinese government agencies give three different estimates of of installed solar PV capacity in China at the end of 2012 (7,970, 7,000 and 3,500mWp) and four different estimates of the amount of solar PV capacity added during 2012 (5,040, 4,500, 3,500 and 1,090mWp).
  • According to the BP 2013 Statistical Review solar PV generation in Canada in 2012 was 0.7 terawatt-hours and according to Observ’ER it was 0.262 terawatt-hours. According to BP solar PV generation in Australia was 2.8 terawatt-hours and according to Observ’ER it was 1.483 terawatt-hours.

And the reason different data sources give such different estimates is that national solar PV statistics are basically guesses. There are too many solar PV systems coming on line too quickly (one every four minutes in the US) to keep proper track of installed capacity and it’s impossible to compile accurate generation totals when output from most PV systems is unmonitored. So different people take their best guesses and come up with totally different answers.

Which means that we can’t calculate meaningful solar PV load factors from national installed solar PV capacity and generation statistics. So assuming it would be nice to get some meaningful numbers, how do we get them?

Here I do it by calculating load factors for individual solar PV systems for which monitored output data are available and by using the arithmetic mean of the load factors to define the average load factor for the country. (I use arithmetic means because solar PV load factors don’t change with the size of the system, all other things being equal, and because production-weighted means often lead to a large number of small systems getting swamped by one large one.)

The monitored output data I used comes from two sources:

  1. gives data for approximately 450 solar PV systems in different parts of the world (although mostly in Europe and the US) ranging in size from 200 watts to 13 megawatts. In roughly half the cases there are enough metered output data to allow load factors to be estimated with good confidence.
  1. lists approximately 40,000 solar PV systems, with the major contributors being Germany (13,637), US (4,847), Australia (2,821), Netherlands (2,523) and France (1,523). Many of the systems, however, contain no usable monitored data or are too recent to have completed a year of operation.

I calculated load factors for approximately 200 systems from data set 1 using a minimum of one complete year of generation data (some systems gave as much as five years). Working through all the entries in data set 2 would undoubtedly have yielded some useful results but would have taken years, so I confined my activities to calculating load factors for an additional ~200 systems from this data set to fill in gaps in the data set 1 coverage.

Figure 2 illustrates the type of metered output data provided, although it’s not always this complete. The data can also be downloaded in numerical form, usually to fractions of a kWh:

Figure 2: Metered generation data. Slepe Farm solar PV system, Dorset

With data like these load factors can be calculated precisely (the example above gives 12.43% in 2012 and 12.48% in 2013 at the listed system capacity of 492kWp).

Figure 3 plots all ~400 (actually 4o9) of my load factor estimates against latitude:

Figure 3: Load factors at 409 monitored solar PV systems

The results are a good deal more plausible than those shown in Figure 1. Load factors in the Northern Hemisphere peak somewhere between 15 and 35 degrees latitude, i.e. in desert areas where solar irradiance is highest, and decline rapidly towards the Pole (decreasing solar irradiance and increasing cloud cover) and more slowly towards the Equator (increasing cloud cover overcomes increasing solar irradiance). The pattern in the Southern Hemisphere is similar. The scatter around the mean (red line) can be attributed to differences in the type and quality of the installation – tracking arrays, optimally-aligned fixed panels, non-optimally aligned fixed panels etc. and to variations in cloudiness between countries/regions at the same latitude. (Tracking arrays are not always identified in the data sets but they invariably show appreciably higher load factors in cases where they can be identified. The point at 65N that gives a 14.1% load factor is a tracking array, as are at least some of the points that give the plus 20% load factors between 25N and 41N.)

Figure 4 plots average load factors against latitude for the countries and regions (six in the US and three in Australia) where I had enough data to calculate load factors with reasonable confidence:

Figure 4: Average load factors by country/region versus latitude

The results are remarkably consistent (note how the Northern and Southern Hemisphere load factors mirror each other), and with a few exceptions the scatter can be explained by variations in cloudiness. Fewer clouds explain why load factors are higher in the Southwest USA and Israel than in Japan and China and higher in Spain, Portugal and Greece than in the Northeast USA. Fewer clouds in summer also explain why the Baltic region has a higher average load factor than the UK even though it is seven degrees farther north, and also why it has a significantly higher average load factor than Norway, which is at the same latitude. And the low load factors in Malaysia and Indonesia? These countries are in the intertropical convergence zone, which is arguably even cloudier than Scotland.

However, not all the differences in load factors can be explained by clouds. Average load factors for Alaska, Spain, Portugal and Greece are likely biased high relative to the other countries/regions by a disproportionate number of tracking arrays (the average load factor for Alaska decreases from 10.1% to 8.9% when tracking arrays are deleted) and the somewhat low average load factor for China may be at least partly a result of unexplained gaps in the records, examples of which are shown in Figure 5.

Figure 5: Gaps in Chinese records

Table 1 summarizes the Figure 4 results with countries/regions sorted by decreasing load factor. The champion is the US Desert Southwest, which will come as no surprise to anyone who has lived there. However, there are as yet no monitored solar PV systems in the Sahara, or at least none for which I have data.

World’s southernmost solar PV system – Ushuaia, Argentina. Latitude 54.8S, load factor unknown

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

33 Responses to Estimating Global Solar PV Load Factors

  1. Euan Mearns says:

    Roger, thanks for excellent post. I have done coding in a bit of a hurry under the residual influence of biofuel. Hope all is OK, changed longitude to latitude.

    This chart shows variations in global cloud cover with time 1983 to left, 2008 to right. It is based on global D2 cloud data I downloaded from NASA and the chart is made using software that NASA provides. Made it about a year ago. I guess the Southern Ocean would be a bad place for floating PV 😉

    I’m out all day and away for the weekend again.

    • Roger Andrews says:

      Hi Euan:

      Thanks for the cloud plot. Here’s one with Europe load factors superimposed on total horizontal solar irradiance, which takes both solar radiation and clouds into account (I added Poland and Romania to fill in the gaps). Things seem to fit together quite well:

      I did a little work on Scotland too. Load factors turned out to be higher than I thought they were going to be, with an average of 9.9, and there’s a system in Aberdeen with a load factor of 9.93. So – what are you waiting for? 😉

  2. Graham Palmer says:

    Roger, great post and data collection.

    There is some good international PV output data at PVOutput.orgI’ve discussed this in another thread – a lot of the data is not available or simply lost because of PV self-consumption, and different jurisdictions use different metering. In Australia, estimates are based on post-code (zip code) data for solar rebates. This provides a good first-estimate but doesn’t allow for panel orientation, dust, shading or failure etc. The Australian PV Institute uses the link above to calibrate estimates from actual data. So at least they are trying to get better data. Moving towards a properly quality-controlled system and knowing exactly what most systems are doing is going to be a lot harder. The systems being logged on this site are probably biased towards enthusiasts and therefore might be better quality and maintained systems.

    • Roger Andrews says:


      Thanks for the links to the data sets, but they seem to present only short-term results. Are there any long-term data buried in there?

      Regarding your comment about monitored systems being biased towards enthusiasts, the only way I can think of checking this is to plot load factors against system capacity, which covers the range of enthusiasm from dedicated homeowners to hard-nosed utility-scale generators. The plots, however, show flat lines all the way. (I should probably have put these graphs in the text):

      Here’s a picture of an installation run by enthusiasts – the solar building in Anchorage. Alaska, load factor 7.7%. At least the panels point south.

      • Re Alaska, vertical panels – I’m trying to think about this, good for winter sun but it’s going to miss a lot of the summer sun when most of the energy should be generated. Only 7.7% – who cares? At least it looks good.

        Re, I haven’t used this for data but clicking on the second column takes you into the stats for the particular installation and the “kWh/kW” is reported.

        Re monitored systems, I was more thinking that as systems age, what is the upkeep going to be on the millions of systems installed in recent years? Will householders pay $1,000 to replace the out-of-warranty inverter, get panels looked at when there is a cell or diode failure, clean panels, ensure shade-free operation as trees grow, etc and the solar company has gone broke? Unlike, say cars or washing machines, life still goes on with a broken grid-connect solar system. Houses will get sold and new owners might not be interested. Assessing this in 10 years will be an interesting research exercise. I’m assuming that enthusiasts will be more interested in keeping systems clean and functional.

      • Euan Mearns says:

        So this array will unlikely ever recover the energy used to create it and since the CO2 used to create is already up there working its evil, this type of renewable installation is a major cause of accelerating emissions and any global warming that they may cause.

      • Roger Andrews says:

        Some Alaskans are REALLY enthusiastic about solar:

  3. Graham Palmer says:

    When I was researching the abatement cost of PV, I wondered about the rebound effect of generous feed-in tariffs (i.e the feed-in tariff would create a disincentive to conserve energy and therefore offset some of the gains of PV). Social theory might suggest that PV households are also efficiency/conservation minded, but economics says that a net-reduction in energy cost will lead to greater energy consumption. Data collection was going to be hard, and there really needs to be sufficient rooftop PV systems running over a few years to assess this.

    Well it turns out that an Australian network operator has recently looked at this (see Mike Swanston starting page 6). Australian electricity prices have risen significantly in recent years, mostly due to rising network costs, but also FiT’s, carbon charges, RET etc. This has resulted in a sharp moderation in demand, which is also driven by offshoring of energy intensive manufacturing, etc.

    In the network area of interest, households without PV have responded to rising costs by moderating consumption by 16% through efficiency and conservation. But households with PV haven’t moderated at all – so this seems to be a sort of Jevon’s Paradox applied to PV. Actually this is just Economics 101 – we live in a strange world.

  4. Kit P says:

    Solar PV does work. Mathematically, CF = 0. CF is the measured power production compared to the rated power. Not the estimated or projected power production. Those are terms that the scam artists who sell junk use.
    I have seen utility scale solar in the US Southwest get 19% compared to the ‘expected 21%. Then some bean counter figures out that the cost of maintenance is more than the value of the power produced. After a few years, power production =0.

  5. Pingback: Links 21st June 2014 | AQONEMAKI

  6. Kit P says:

    Graham missed an important failure mechanism. One of the names I have for PV systems is smoke emitting diodes. I am a trained fire fighter and in electrical safety, CPR, and first aid. During my last year in the we had three electrical fires. Two nuclear reactors, lots of high explosives, and metal hulls sitting in salt water changes how you thing about electricity.
    There have been a significant number of house fires started by PV systems. You cannot put out a fire with water until electricity is de-energized. This begs the question of how you stop a PV panel from making power.

    • Graham Palmer says:

      Kit, this is from The Australian 19 May 2014.

      A Queensland company that sold allegedly faulty circuit breakers that caused at least 70 burnouts in rooftop solar panel arrays has gone bust, leaving tens of thousands of homeowners at risk of electrical fires.

      Advancetech, based on the Sunshine Coast, went into receivership on Friday, only four days after Queensland Attorney-­General Jarrod Bleijie ordered the immediate recall of 27,600 Avanco-branded DC solar power isolators imported and sold by the company.

    • Ed says:

      Surely building regulations must stipulate inclusion of one or more manual insulation switches, one of which must be accessible from outside the building ?

      • Graham Palmer says:

        Ed, the isolator is located on the roof near the panels and Australian installations must comply with AS/NZS 5033. The problem is that –

        a) it’s a bit hard to get on the roof via a ladder to cut wires or disconnect terminals when there’s a fire
        b) the isolator is hard to switch off when its the thing that’s caused the fire and probably a blob of burning plastic
        c) a hose is not wise and few householders would have a dry chemical or CO2 fire extinguisher on hand
        d) it’s sunny and there’s several hundred volts DC being generated by the panel string

  7. Kit P says:

    Graham there are two kinds of electrical devices. Those that have failed and those that have yet to fail. In the article you linked no information was provided to say why they were faulty. If you have 27,000, 70 failures may be the expected failure rate. This is why electricians at power plants wear protective equipment.
    We take risks in life because there is a benfit. For there to be a benefit of PV, the risk must be minimized while the production maximized. So far I find no evidence that this is happening with PV. Lot of evidence that PV is a scam.

  8. Ed says:

    Can someone help me with some basic maths. 10 kWp system with a load factor of 10% will give you an average 1kW of power. That is 24kWh of energy per day. Correct?

    • Averaged over a year, yes.

    • Euan Mearns says:

      To clarify Roger’s response. Some days the load may be 1% (Raining in Aberdeen in December) and other days it might be 20% (sunny in Aberdeen in July). The average load factor is integrated over 365.25 days. An important point to take away from this is that solar load is negatively correlated with seasonal demand.

  9. Ed says:

    What is needed is a cheap battery technology to be developed. One promising project is the organic megaflow battery from Harvard university.

    • Roger Andrews says:

      We needed to develop a cheap utility-scale power storage technology BEFORE we started to spend gigabucks on non-dispatchable wind and solar, not after.

      • Ed says:

        The vast majority (>90%) of our money (energy) is being squandered on fossil fuel infrastructure which will be redundant in a few decade’s time. However limited renewable energy is, and I admit there are problems to be overcome, it is got to be better than building more airports, more roads, more weapons of mass destruction and growing our population.

  10. Pedro A. Prieto says:

    Euan, it is difficult to estimate load factors per country with the BP data. It is a way to failure. Spain, for instance, has more than 20% of installed power than it is accounted officialy by the Spanish government and by BP that has probably taken data from the Spanish government. This type of overpowering the installations was due to the laws attributing the best given feed-in tariff to plants below 100 kW. Then according to later definitions, this limit was considered to be at the output of the inverter. Therefore, many promoters installed from 105 to even 130 kWp power at modules level, because inverters were able to cut the output to 100 kW, but they were giving more peak hours a year. In our book (Prieto et Hall) on solar PV EROI, we were very conservative and calculated an 8% overpowering for this concept. Recently, the industry admitted a 15% when the government started to measure and assess the solar PV plants by the installed peak power, thus reducing the income by the originally granted FIT.

  11. kakatoa says:

    Andrew- it looks like our last posts were lost when the host changed. My original message is located at the end of this one in case you didn’t save the link….

    Responding to your response to my systems output-
    Yes my PV system, located off the Highway 50 corridor halfway between Sacramento and Lake Tahoe has held up rather well over the years- output wise. Back when I put my system in place the CEC estimated it’s yearly output at 9380 Kwh. My cash outlay was $44K before rebates ($14400) and tax credits ($2000). My 12 and 24 month rolling average yearly kWh output values are from a separate (not my inverters total kWh recorder) meter I had put in place 6 six ago. My inverter’s total kWh values are 4.8% higher than the independent meter I have in place.

    Original post is below:

    I wish the roof report, see link above, would of documented who would remove and reinstall a 5kW PV system for $500! The $7000. reference for the task would lead me to going off our roof to a ground mount system! Thought you might like the reference for your levelized costs files. Pg. 17 notes this:

    …”That said, solar installers and SFCs’ cost
    estimates to remove and reinstall a 5‐kW system ranged widely, from as little as $500 to as much as
    $7,000 for an asphalt shingle roof, with a median expected cost of $2,750 (estimates for a concrete/clay
    tile roof are slightly higher).”….

    A few assumptions are made later in the document to come up with some cost estimates to add to roof mounted PV systems levelized costs…..

    My little PV system just hit it’s 8th year of operation. It was rated by the CEC at 6.12 kW DC and 5.22 kW AC. My 12 month rolling average output is 9326 kWh. The 24 month rolling average is 9400 kWh.

  12. Roger Andrews says:

    Kakatoa; I don’t remember exactly what I said but to complete the record here goes again with a few additions.

    Thanks for the info on your system. According to your numbers it has a load factor of 17.5%, which seems about right relative to my average for California (16.9%) and your location, which I guess would be a couple of thousand feet above sea level somewhere around Placerville, correct?

    Your installation costs work out to $7,200/kWp installed before rebates, which I don’t think is all that much higher than present-day California costs despite the fall in the price of solar panels.

    Your output also doesn’t seem to have deteriorated any in eight years of operation. I’ve been assuming in my economic runs that solar panel output decreases by about 1% for each year of operation. Guess I may have to rethink that.

    Thanks for the CSI links. I hadn’t realized that roof composition was such a big deal north of the border. All the roofs down here are flat and made of reinforced concrete, so you can’t replace them, and sooner or later they all leak anyway.

  13. kakatoa says:


    My system is degrading a bit so your 1% loss is likely ok. My swag is that it is down about 4% in the last 8 years. My Mitsubishi 170 watt panels were built with the last of the super quality silicon back in 2006. The output out of the box was a bit higher than rated as my first two years of output was closer to 9800 kWh. I was able to keep my costs down back in 2006 by being the general contractor, I selected the hardware from a wholesaler and had a small firm install the system on the roofs I wanted them on. I saved a few bucks taking care of the permitting paperwork.

    I needed to check the CA solar data base today to see if a quote for a 25.7 kW system was reasonable for a new Boys and Girls club a local community group is finalizing the plans for. It is a bit amazing how far the prices have dropped for PV systems. The quote was for $3.50 watt (DC) installed. The last quarter of 2013 data, for PG&E’s service territory >10kW systems indicated costs of $4.79 watt for residential systems and $3.29 watt for commercial installs.

    Given the recent tariff increases for PV panels from China the quote seems reasonable. We need some more details about the buildings specific energy needs and the possible rate schedules before we finalize if the system is sized ideally…..etc..

  14. Kit P says:

    It appears my response was lost too. Kit is living in China at the moment. When I lived in California 25 years, I put the best tile roof on the dream house I built. I also designed a solar hot water that was mounted on the south facing hill. Just off CA highway 88 at the 2000 ft elevation. Solar energy collection is an interesting hobby but very expensive hobby.
    Kakatoa has an interesting hobby too. His system makes about $500/yr on the market. A very expensive hobby but the longer his systems last the less expensive his hobby will be. Since Kakatoa installed an independent meter that tells me that he is one of the rare individuals that will keep a system working.
    The problem with silly statistics is the realities of life. Kakatoa will have to decide what to do based on the actual cost of a repair. For every large nuke plant you will need 2 million like Kakatoa living at 2000 ft elevation. Let me start the count. I am one, Kakatoa is two. BTW, we have to keep living there 60 years after installing the system.

    • Mark says:

      Hi Kit,

      I concur with you that not a lot of folks who invest in PV have a clue about how to keep the system operating to it’s potential! From what I have read this is really a problem with some early adaptors in the San Diego area- especially in the public sector. We live on 11 acres without city services so by necessity we have had to become jack’s of all trades to some degree.

      I also concur with you, at least until CA decided to fiddle with the residential rates AND force the big three to get 20% (now 33%) of their generation from RE no matter what it costs them a few years back, that the wholesale market value of the output of my system would only be worth give or take $500.00- ie 5 cents a kWh.

      Amazingly your estimate is almost the exact same value as the CPUC said a kWh of RE cost the big three to source RE Generation way back when. “From 2003 to 2011, contract costs have increased from 5.4 cents to 13.3 cents per kWh.” Somewhere on my old PC I have the specific references for the quote above if you would like it. During this same time period the contract for the long term PPA for the state’s first concentrating solar project was approved. It came in at something closer to 17 cents a kWh. As an FYI the PPA’s have “Time of Delivery- TOD” adjustments “Actual RE payments are adjusted by each IOU’s individual TOD factors and the time that a project generates electricity…”

      All the fiddling of the residential rates has led to our rather unsustainable very progressively tiered residential rate structure. Depending on how you look at things the value of my output these days would be as low $.09 kWh to up to as much as $0.36 kWh. A few specifics of my usage AND WHEN I send some 240 voltage, at least it used to be 240 a few years back, but it appear PG&E has upped their voltage to 251/2 at my location for the last few years, watts to the grid are needed to figure out the value of a kWh from my system. By 2022, when the 33%RES is meet by PG&E, the value of a kWh from my system is going to be more than a bit up in the air. The current residential rate design leads to this problem: “Absent any change in the residential rate design methodology, the difference between Tier 2 and Tier 4, which was 18.9 cents in Jan 2012 (33.5 vs 14.6 cents per kWh), is forecasted to increase by 65 percent to 31.1 cents in 2022 (50.5 cents vs 19.4 cents per kWh).” Ref: PG&E 2012 Rate Design Window 2012 Application.

      Your welcome to stop in at our place if your going to be in the Sierra’s. We are at an elevation of 2400 feet so we get all four seasons. We don’t live to far from a few of PG&E’s hydro facilities on the American River. I guess some physics would lead to our grid supplied electrical energy coming from those facilities rather than the utility scale PV farms down in the deserts a few hundred miles south of us. It’s been close to 20 years since I was in China, a lot of bicycles were used back then for transportation around Beijing. Similar to what my farther said it was like in 1974/5 when he was there for a tech transfer effort after H. Kissinger and president Nixon improved relations with China. He worked for an evil (/sarc) Fossil Fuel company that was rather well known at the turn of the last century. I don’t think it was his former firms activities that caused the Cuyahoga River to burn.

  15. Kit P says:

    Have we met, I was at the A’s game rearing a tribe hat. My father grew up swimming in the Cuyahoga River on his grandfather’s farm. I was born in a small town near there and my father was very bitter about how the environment was impacted. The irony is that he moved to the Santa Clara Valley before it was ruined. I do not need to image it as it was described by Jack London in CALL OF THE WILD.
    I moved back to California to work for SMUD at Rancho Seco because we could live in the Sierra’s with many acres to ourselves. Politics made my job go away. The nature of my job allows me to see lots of different utilities. Some are well managed and poorly managed. My job is a lot easier and more enjoyable at the well managed places. California is a model of how not to do things.
    I now work in China. It will likely be my last job before retirement. I am here because at the moment I am enjoying watching and listening to the waves roll in from my air conditioned apartment. Soon I will walk to work. While I have only visited a few places, China has discovered the internal combustion engine. There are still a few riding bikes but there is a plague of small motor cycles. The roads are choked with cars. While I have a choice to live in a big city, the boondock of China are just fine.

  16. Kit P says:

    Euan, China is generating a whole host of internet computer mysteries. I was a little peeved with the direction to a new hotel that showed it in the ocean. Of course just a couple years ago, it was in the ocean before it was filled in. Today my wife informs me that in our junk mail folders was a very nice email from Pear telling us the hotel had a shuttle and where it would pick us up. Candy was also very helpful too. More interesting than fun with excel.

  17. Pingback: Estimating Global Solar PV Load Factors | Energy Matters | | USASOLARREBATES.COMUSASOLARREBATES.COM

Comments are closed.