Photovoltaic is an Energy Source, not a Sink!

This is a comment by Luis De Souza on a recent paper by Ferroni and Hopkirk who reported a negative energy yield for photovoltaic plants in Switzerland (in other words, an energy return, EROEI, smaller than one). It is an anomalous result, considering that a comprehensive meta-analysis of the field reported values of 11-12 for the EROEI of the most common PV technology. So, what's wrong with the paper by Ferroni and Hopkirk? A lot of things, it seems. Here, De Souza shows that photovoltaics is a source of energy, even in a not so sunny country as Switzerland. He concludes that something went badly wrong with the review procedure with the journal that published the paper by F&H, "Energy Policy". That seems to be correct and you may be interested to know that an extensive rebuttal of that paper has been prepared and submitted to the journal by a group of researchers expert in the field of energy calculations. That rebuttal finds a lot more wrong things in F&H's paper than those identified by De Souza. In short, Energy Policy managed to publish a flawed study that should never have been published in a scientific journal. Unfortunately, it was done and now a lot of people are using it to support the war against renewable energy.

Photo-Voltaics is not an energy sink in Switzerland

by Luis De Souza

Energy Policy recently published a study conducted on the EROEI of Photo-Voltaics (PV) technologies installed in Switzerland. The end result is a remarkably low figure of 0.8:1, well below any EROEI assessments ever conducted on this energy technology.

Such a figure naturally made the delight of those campaigning against renewable energy, who take at face value any hints of negative performance. However, from this study a number immediately stands out: average lifetime energy yield of 106 kWh/m2/a. As it turns out, a closer look at this single figure is enough to disprove the hypothesis of PV being an energy sink in Switzerland.

Basic check

The first check one can conduct on this EROEI study is to compare it with previous assessments. Pedro Prieto and Charles Hall produced what is possibly the most conservative EROEI study on PV, concluding on a figure of 2.4:1 for Spain. There is much to question in this study, in particularly the arbitrary translation of non physical requirements of a PV system into energy inputs, but for the purpose of comparison let this low figure be taken at face value.

Yearly solar radiation at the latitude of Madrid (40 ºN) is in the range of 2 000 kWh/m2. At the latitude of Bern (47 ºN) this value is down to 1 500 kWh/m2. Assuming the extraordinarily high energy inputs computed by Prieto and Hall for Spain also apply to Switzerland one can directly apply the rule of three to compute an EROEI figure of 1.8:1.

Mind here that EROEI is a logarithmic measurement. Therefore 1.8:1 is considerably closer to 2.4:1 than to 0.8:1. These simple figures start showing that something is fundamentally awkward with the results presented by Ferroni & Hopkirk.

Why energy per unit of area?

The article in itself is not very detailed and leaves much for the reader to guess. However, there is a key figure that plays into this EROEI study that immediately stands out: an average lifetime energy output of 106 kWh/m2/a for solar panels installed in Switzerland. Upfront, it appears a strangely low figure, but there is something more problematic with it. Each solar panel model is designed and built differently, with cells distributed in different ways; even among those produced by the same manufacturer the capacities per unit of area can be quite different.

The graph below shows capacities per unit of area for different models presently on sale by various manufacturers, including the world's top three.

While Ferroni & Hopkirk never indicate what energy output per installed capacity they use, this sample of panel capacity per unit area allows for some investigation into it. The figure below presents this calculation for these same panel models. 

Again, the figures vary widely, with the average under 700 Wh/Wp/a.


Comparison with PVGIS

PVGIS is a web application developed by the Joint Research Centre (JRC) that calculates the energy output of a PV system taking into account yearly solar insulation, panel orientation and system losses to cabling, the inverter, temperature, angular reflectance and more. PVGIS has not been updated in a few years and for the most recent systems I have been involved with it underestimates first year output by 5% to 10%. But for this exercise its results are taken at face value.

The table below is the result produced by PVGIS for an hypothetical system rated at 1 kWp, optimally oriented and installed around where I live, in the Canton of Zürich (47 ºN, in the Northwest of Switzerland). The most relevant figure in this report is the energy output estimate: 1090 Wh/Wp/a. While this is an estimate for an optimally oriented system, it provides a good measure of where the annual energy yield figure used by Ferroni & Hopkirk actually lays. 

PVGIS © European Communities, 2001-2012
Reproduction is authorised, provided the source is acknowledged
See the disclaimer here


Comparison with Swiss statistics

Ferroni & Hopkirk cite the statistics compiled by Swiss Federal Office of Energy (SFOE) as the source of their 106 kWh/m2/a figure. There are a number of different documents available from theSFOE website covering all matters of energy generation and consumption.

In recent years the SFOE has produced a yearly report of renewable energy with a series of important figures. The report for 2015 is not available yet, therefore the figures used here refer only up to 2014. These are all aggregate values, but are already enough to provide another investigation path into Ferroni & Hopkirk's figure.

After going through these reports, one thing becomes evident: the SFOE does not use the energy output per unit area measure cited by Ferroni & Hopkirk. As expected, average electricity generation figures are rather provided in energy output per installed capacity (Wh/Wp/a).

Secondly it is important to note that PV is something relatively new in Switzerland, installed capacity has picked up only recently, almost tripling from 2012 to 2014. At the end of 2014 there were 1060 MWp of PV panels installed in Switzerland, a figure that grew 40% that year alone. During 2014 electricity generation from PV reached 841 GWh.

Assuming that all the new systems installed in 2014 were connected to the grid on the 1st of January a figure 794 Wh/Wp comes out for the year. This is already on the high side of the possible generation per installed capacity figures used by Ferroni & Hopkirk. However, assuming that these new systems where connected to the grid at a regular pace throughout the year, this number rises to 927 Wh/Wp. This is less than 15% off the PVGIS estimate, and possibly explainable by non optimal orientation of some systems and a small fraction of older and likely less efficient systems. Usually, systems tend to be installed towards the end of the year, to take up the most favourable legislative framework.

Possible causes

The first cause that comes to mind for such low energy yield figure is an erroneous cell efficiency factor. PV cells are rated in control experiments where their energy output is assessed at a temperature of 25 ºC and a constant radiation of 1 kW/m2. This assessment is very useful to compare different cell technologies. Modern day wholesale crystalline cells reach efficiency factors between 14% and 16%, i.e. they convert that fraction of incident radiation into electrical current.

Since Ferroni & Hopkirk present average lifetime yield in energy per unit area, these authors might have converted incident radiation in Switzerland directly into an energy yield. However, instead of using the figures above, the efficiency factor they used must have been in the order of 8% to 9% to result in an energy per installed capacity value around 690 Wh/Wp/a. Such low conversion factors are more common with thin film technologies.

A second hypothesis is the employment of an unusually high cell degradation rate. PV cells loose their properties over time, both to the heat they are exposed to, as to the solar radiation itself. While tools such as PVGIS can easily model system losses, they usually leave this degradation rate out. Research centres such as the JRC have assessed PV technologies for decades, concluding on an energy yield degradation rate in the order of 0.5 %/a. Moreover, these long term studies also indicate that cells tend to degrade in a linear fashion.

The following figure presents two hypothetical degradation rates that bring down a PV panel from 1090 Wh/Wp/a to an average yield of 690 Wh/Wp/a over a 25 year lifetime: a liner degradation of 33.5 Wh/Wp/a and a logarithm decline of 4 %/a. In both cases the energy yield dives under half before the end of system life.

While this latter hypothesis is my favourite, it does not explain the employment of the strange energy per unit area figure. Also, these degradation rates would assume that in the face of a fast collapse in energy output owners would never activate panel warranty. 

Final remarks

Replacing the inexplicably low energy yield figures used in this study by those available from the SFOE is already enough to bring the Swiss PV park into positive net energy territory. However, such result is still far from previous PV EROEI assessments, even the highly conservative estimate produced by Prieto & Hall. Just as the energy yield assumptions proved problematic in this study, I expect similar awkwardness to be found on the energy input side of the equation. However, I leave this aspect to be assessed by someone else.

The publication of such a study by a relatively renowned outlet begs for deep reflection. The last article I authored in a scientific journal was over two years in review; this is usually a slow and painstaking process. Being myself an editor and reviewer at scientific publications, I am at a loss to explain how could such a problematic figure of 106 kWh/m2/a have possibly made through the peer review process. It should have immediately raised a red flag to whoever is slightly acquainted with PV technology and economics, calling for close scrutiny by reviewers and editors alike. Something fundamental has failed in the review process at Energy Policy. 

The Take Away

The EROEI figure concluded by Ferroni & Kopkirk for PV is the lowest ever and far below any previous studies.

These authors use awkward units that largely obfuscate their assumptions on yearly energy yield.

A sample of various panel models points to an energy yield under 700 Wh/Wp/a used in this study.

Official statistics point to an average yield well above 900 Wh/Wp/a for the Swiss PV park; this is in line with values from assessment tools like PVGIS.

The peer review process is not functioning properly at Energy Policy.

But what's the REAL energy return of photovoltaic energy?

According to a recent, comprehensive study of the scientific literature (1), the average energy return on energy invested (EROEI) of the most common photovoltaic technology (polycrystalline Si) is 11-12. A far cry from the legend of the "EROI smaller than one" that's making the rounds in the Web

Some time ago, a colleague of mine told me the story of when he had been in charge of the installation of one of the first photovoltaic plants in Italy, in 1984 (shown in the figure, on the right). He told me that, shortly after the installation, a high-ranking politician came to visit the plant. As a demonstration, my colleague connected the plant output to an electric heater, lighting up the internal heating elements.

The politician refused to believe that the heater was being powered by the PV plant. "There has to be a trick," he said, "this is not possible. It must be a scam." My colleague tried to describe to him how PV cells work, but imagine trying to explain quantum mechanics to a politician! Apparently, he left still unconvinced.

More than 30 years have passed from the installation of that old plant, but the general attitude about photovoltaic energy doesn't seem to have changed a lot. Not that people think that photovoltaic is a scientific hoax in the same league as the many proposals about such things as "free energy" or "cold fusion" (or maybe yes). But it seems that many people just can't believe that those small blue things can produce energy in any significant amount. Come on: in order to produce energy you need an engine, a boiler, a smokestack, a turbine, something like that.....

Indeed, most of the current discussions on photovoltaic energy seem to turn around one or another kind of legend. The most recent one seems to be that photovoltaics has a low energy return (EROI or EROEI), sometimes said to be even smaller than one. If it were true, it would mean that photovoltaic plants are not producing energy, they are just consuming it! But it is not true. It is just one more example of confirmation bias: cherry-picking the data that confirm one's pre-conceived ideas.

It is true that you can find a few studies (very few) that look serious (perhaps) and that maintain that PV has a low EROI. However, in a recent study, Bhandari et al. (1)⁠ surveyed 231 articles on photovoltaic technologies, finding that, under average Southern European irradiation, the mean EROI of the most common PV technology (polycrystalline Si) is about 11-12. Other technologies (e.g. CdTe) were found to have even better EROIs. Maybe these values are still lower than those of some fossil fuels, but surely not much lower (if they are lower) and a far cry from the legend of the "EROI smaller than one" that's making the rounds on the Web.

Then, if you are worried about another common legend, the one that says that PV cells degrade rapidly, think that those of the plant described at the beginning of this article were found to be still working after 30 years of operation, having lost just about 10% of their initial efficiency! In addition, consider that the most common kind of cells use only common elements of the earth's crust: silicon and aluminum (and a little silver, but that's not essential). What more can you ask from a technology that's efficient, sustainable, and long lasting?

All that doesn't mean that a world powered by renewable energy will come for free. On the contrary, it will take a very large financial effort if we want to create it before it is too late to avoid a climate disaster (quantitative calculations here). But a better world is possible if we really want it.




(1) Bhandari, Khagendra P., Jennifer M. Collier, Randy J. Ellingson, and Defne S. Apul. 2015. “Energy Payback Time (EPBT) and Energy Return on Energy Invested (EROI) of Solar Photovoltaic Systems: A Systematic Review and Meta-Analysis.” Renewable and Sustainable Energy Reviews 47 (July): 133–41. doi:10.1016/j.rser.2015.02.057. 

Here is the relevant figure from the article:

h/t Domenico Coiante, Marco Raugei, and Sgouris Sgouridis

An energy miracle? But we already have it!

Silicon is a material with properties close to the optimal for a solar cell. It is also one of the most abundant elements in the earth's crust, and, finally, we know how to use it to manufacture cells with efficiency close to the theoretical maximum. Isn't it a miracle?

"EnergySkeptic" recently commented on an article appeared in "Nature" in 2014 on the possibility of cheap photovoltaic cells entering the market of solar energy. The post is short enough that I can reproduce it in full, below. It is interesting because it shows the problems with the idea of the "miracle breakthrough" in energy that Bill Gates advocates.

Here, the discussion is on perovskite solar cells; a technology that promises to be cheaper than that based on silicon. Perovskites are a large class of materials; those being studied as solar cell materials have several advantages, including the fact that they can be manufactured in the form of thin films, don't need to be so extremely pure as silicon, have a band gap close to the theoretical optimum.

That, however, doesn't necessarily make perovskites a "breakthrough" in the field. Even assuming that perovskite cells could reach an efficiency high enough to be marketable, the problem is that, at present, the cost of the cells is only about 30% of the total cost of a solar plant. Even if perovskite cells were to cost half as much in comparison to silicon ones, that would be no improvement unless their efficiency were to match or exceed that of silicon. Otherwise, the whole plant would probably cost more because it would have to occupy more space.

In practice, to have a breakthrough in solar power, we would need a technology which is 1) significantly cheaper than silicon, 2) much more efficient, 3) that uses no rare and non-renewable elements (that rules out, in the long run, cells that use tellurium or gallium). That's a tall order, especially considering that we are bumping into the physical limits of single-junction cells; which cannot have efficiencies higher than a little more than 30%. Silicon, because of some quirks of the way the universe works, happens to be placed almost in an optimal position in terms of band-gap and, at the same time, to be a widely available element in the earth's crust. So, it is, in many respects, an optimal choice for solar cells, and already not so far away from its theoretical limits. I think that we'll stay with silicon for a long, long time. Surely we will improve the technology, but don't expect miracles. That silicon works so well is already a miracle!

_______________________

Further notes:

1. Here, in Florence, a colleague of mine has built a nice solar plant that uses multi-junction GaAs cells and concentrating mirrors, attaining, I think, around 50% efficiency. I saw it: it is a wonder of technology, full of gears, motors, optics, sensors, computers, and things. But I didn't dare to ask him how much it would cost to buy one for the roof of my house!

2. True breakthroughs may occur "downstream" with respect to energy production; for instance with batteries and the diffusion of a new generation of electric vehicles. There is no thermodynamic limit to the number of times that a battery can be recharged without degrading.

3. "heavy-duty trucks, locomotives, and ships run on diesel fuel" in the article below is, in part, a canard. Here in Europe, locomotives already run on electricity. Trucks can run on electricity, too, (http://mondoelettrico.blogspot.it/2014/08/ehighway-il-filocarro-elettrico.html). For ships, the problem is not so much how to push them on, there are ways. It is another one, much more difficult (see e.g. https://blogdredd.blogspot.it/2015/08/why-sea-level-rise-may-be-greatest.html). And the only way to solve that problem is to rush into renewable energy as fast as possible.


_________________________

Van Noorden, R. September 24, 2014. Cheap solar cells tempt businesses. Nature #513 470-471.

[Excerpts. Of interest because rarely do obstacles get mentioned in the news. Most are optimistic hype making it sound like a solution to the energy crisis is just around the corner. And forget that electricity does not solve our main problem — heavy-duty trucks, locomotives, and ships run on diesel fuel ]

Large, commercial silicon modules convert 17–25% of solar radiation into electricity, and much smaller perovskite cells have already reached a widely reproduced rate of 16–18% in the lab — occasionally spiking higher.

The cells, composed of perovskite film sandwiched between conducting layers, are still about the size of postage stamps. To be practical, they must be scaled up, which causes efficiency to drop. Seok says that he has achieved 12% efficiency with 10 small cells wired together.

Doubts remain over whether the materials can survive for years when exposed to conditions outside the lab, such as humidity, temperature fluctuations and ultraviolet light. Researchers have also reported that ions inside some perovskite structures might shift positions in response to cycles of light and dark, potentially degrading performance.

The need for complex engineering might create another setback, says Arthur Nozik, a chemist at the University of Colorado Boulder. After plummeting in past years, the price of crystalline silicon modules — which make up 90% of the solar-cell market — has leveled off but is expected to keep falling slowly. As a result, most of the cost of today’s photovoltaic systems is not in the material itself, but in the protective glass and wiring, racking, cabling and engineering work.

When all these costs are factored in, perov­skites might save money only if they can overtake silicon in efficiency. In the short term, firms are focusing on depositing the films on silicon wafers, with the perovskites tuned to capture wavelengths of light that silicon does not. On 10 September, Oxford PV announced that it was working with companies to make prototypes of these ‘tandem’ cells by 2015, and that this could boost silicon solar cells’ efficiencies by one-fifth, so that they approach 30%. Malinkiewicz’s hope is to find a niche that silicon cannot fill: ultra-cheap, flexible solar cells that might not last for years, but could be rolled out on roof tiles, or used as a portable back-up power source.

There is another potential snag: perovskites contain a small amount of toxic lead, in a form that would be soluble in any water leaching through the cells’ protection. Although Snaith and others have made films with tin instead, the efficiency of these cells is only just above 6%.