by Rud Istvan and Planning Engineer
There are many journal articles, media stories, NGO papers, and blogs claiming solar already has, or soon will, reach general grid parity. Grid parity is when the cost of solar equals the cost of conventional electricity alternatives. It should also mean equal without subsidies like feed in tariffs (FiT), net metering, and tax credits.
This seemingly simple idea is not so simple. Grid parity depends on what sort of solar, on whose grid, and in what location. Solar needs insolation (sunlight energy), and that is quite variable with latitude and regional cloudiness. Red/orange is good, blue/violet is bad, yellowish is so-so, and greenish is worse than yellow but not as bad as blue.
What might work well in Phoenix will not in Seattle. Or in Germany, which we will examine in more detail, since its grid is now 6.9% solar generation coming from the world’s greatest installed photovoltaic (PV) capacity (now 38.5 GW, 26% of world total PV).
Grid parity is another wicked problem. As with the true cost of wind we are not assigning ‘carbon externality’ costs to conventional generation, which are debatable.
There are two fundamental kinds of solar generation. Photovoltaic (PV) uses semiconducting solar cells to generate DC electricity, which is ‘inverted’ to grid compatible AC. Einstein explained the underlying theoretical photoelectric physics in 1905, which earned his 1921 Nobel Prize. The other kind is concentrated solar power (CSP), which is simply a magnifying glass concentrating sufficient sunlight energy to drive a thermal generator, usually an ordinary steam turbine. There are two sorts of ‘magnifying glasses’: parabolic troughs and heliostat ‘power towers’.
Solar grid parity depends on whether one is comparing residential or utility scale solar and electricity pricing. Residential has the highest grid distribution costs, so the highest electricity price/kWh. A U.S. rooftop PV system cost about $5.80/W in 2014; utility PV cost about $4.5/W (shown below).
Solar parity also depends on the alternative generation mix in the grid comparison, which depends on the grid. We show below that both forms of solar may be at grid parity in northern Chile. Chile electricity generation costs from ~$0.16/kWh (off peak spot market) to ~$0.25/kWh (peak). Chile is split roughly equally into 1/3 inexpensive hydro (in the south), and 1/3 natural gas plus 1/3 coal in the north. Chile imports all of its natural gas (most as expensive LNG), and most of its coal. Northern Chile has inherently high cost generation.
The 2014 Energy Information Agency (EIA) estimates 2019 utility scale solar levelized cost of electricity (LCOE) is nowhere near parity: $130/MWh for PV ($119 after subsidies), and $243/MWh ($224 after subsidies) for CSP. Those generation figures compare to ‘true’ (hidden carbon tax and lifetime adjusted) CCGT or USC coal at ~$57/MWh. The EIA analysis’ table 4 estimates the difference between levelized avoided cost of electricity (LACE, that is, the alternative cost of the actual generating mix in 22 separate US regions) and LCOE. If LACE-LCOE is positive, then solar generation costs less and should be built. EIA says utility PV will be past grid parity by 2019 in some regions (red/orange insolation), but not in most or on average–contrary to Scientific American. CSP will still not be viable in any US regions.
The basic question we explore here is whether, when, and where solar might reasonably be expected to improve enough to reach general grid parity— depending on better efficiency and lower cost. Secondarily, why has so much solar already happened if parity has not yet been reached? And why do the media claim parity has been achieved?
In 1961, Shockley and Queisser (S-Q) first calculated the theoretical quantum efficiency of a simple single junction solar cell. A simple PV cell can be at best ~31% photoelectric efficient converting sunlight energy into electricity.
Theoretical S-Q PV efficiency cannot be achieved in reality. There are a number of hurdles including reflectance, charge carrier recombination, grain boundary charge trapping (in polysilicon), and conductor resistance. These vary by semiconductor material and cell design. For example, the highest efficiency monocrystalline silicon cells add special coatings to reduce reflectance, and use copper rather than aluminum to reduce resistance. They are ‘thick’ to maximize light capture efficiency, but that thickness increases charge carrier recombination and lowers photoelectric efficiency.
Solar cells have been researched for four decades. The three commercial types are maturing in efficiency and in manufacturing cost.
There has been no meaningful increase in research cell silicon efficiency since about 2000 (monocrystalline silicon (mSi) 25%, polycrystalline silicon (pSi) 20.4%). There has been some recent progress in thin film cadmium telluride (CdTe, First Solar 21%). Research cells deceive concerning grid parity. The highest production efficiency mSi panels (uniform dark blue with little aluminum ‘diamonds’ from missing round boule corners) are 21.5% (SunPower). pSi panels (mottled light blue, no ‘diamonds’) from China—which drove the rapid drop in panel prices from about 2010 to about 2013 when China’s SunTech went bust—are now 15.6% (Yingli). First Solar CdTe (black) production panels have reached 14.7%.
On the other hand, high purity high efficiency mSi is expensive. SunPower’s best E20 panels cost about $1.70/W and sell now for about $2.15/W. Thin film CdTe is inexpensive; First Solar says panel manufacturing cost is now $0.54/W. Their panel ‘price’ is under $1/W (uncertain since FSLR sells utility systems, not panels). So PV has high efficiency high cost panels, or low efficiency low cost panels. PV does not have high efficiency low cost panels. FSLR says it can achieve 17.5% CdTe production efficiency by 2017, and EIA says that would enable true utility PV LCOE parity in a few US regions.
As with most ‘factory made’ goods, PV costs follow an experience (aka learning) curve. This is a log/log linear relationship. The PV curve is about 20% cost reduction per doubling of cumulative production. But with so much PV already produced, each additional cumulative doubling takes much longer in calendar time. First Solar says it has now installed 10GW over its entire history. So the next ~20% cost reduction (from ~$0.54/W to ~$0.43) requires production of another 10 GW. Rapid cost reduction lies in the past, not in the PV future.
The reason that PV panel efficiency/cost is such a big deal (now mature and slow to improve on both dimensions) is that panels have become the lesser cost of any PV installation. The remainder is ‘balance of system’ (BOS). BOS includes mounting brackets (and for utility installations, sometimes one or two axis sun tracking mechanisms), wiring, inverters, installation, and maintenance (cleaning dirty panels no different than cleaning dirty windows). At least 80% of BOS costs are directly related to efficiency. For any given amount of generation, a 21.5% efficient panel compared to 14.7% requires (14.7/21.5) ~2/3 the land or rooftop, mounting bracketry, wiring, installation labor, maintenance.
Nobody knows how to significantly reduce BOS other than by higher efficiency (now with no to slow gains). Most BOS costs do not scale, and only inverters follow a learning curve. Grid parity is no longer about PV panels, but rather about BOS. Note that the following 2014 NREL figure is on a WDC basis. The city of Palo Alto says a 5kWDC rooftop installation only produces about 4kWAC (owing to inverter losses, suboptimal roof angle/orientation, …). Note further that NREL BOS ‘reality’ was more expensive than NREL projections for 2012 and 2013. Note that ‘Analyst Expectations’ are those of the PV industry. Finally, note that the actual 2014 residential PV cost in Palo Alto was $5.83/W, not NREL’s ~$3.50/W. California’s 550MW Topaz PV project completed in 2014 cost $1.8 billion or $4.54/W (with capacity factor 23%). California’s 550MW Desert Sunlight PV was completed January 2015 at $4.18/W because it sits on federal land in the Mohave Desert. Those new California utility PVs are presently the largest in the world.
BOS and efficiency are so important that First Solar (CdTe) opened a line in Malaysia in 2014 to manufacture 100MW/year of high efficiency mSi panels in order to compete beyond the utility marketplace.
Given all this information, we return to our initial background question. How did low insolation ‘blue’ Germany become the world leader in installed PV capacity? In short, Germany passed the 2000 Erneuerbare Energien Gesetz (EEG, Renewable Energy Law). Since that time, Germany’s residential electricity price has more than doubled, and Germany now has the second highest rate in Europe after Denmark. Denmark also has a higher renewables penetration (39% wind in 2014) than Germany (19% wind and 7% PV). The EEG Umlage (surcharge for PV FiT) has grown from nothing to €0.0624/kWh, 22% of what a German household paid in 2014. That Umlage provides about €11billion (2014 estimate) of FiT subsidy to PV owners.
Germany’s residential electricity cost is now about €0.29/kWh. When the EEG law passed in 2000, it was €0.14/kWh. Fraunhofer ISE claims residential PV reached German grid parity in 2012. But that is only thanks to Energiewende elevated electricity pricing. It is not close to ‘true’ parity, and probably never could be due to ‘blue’ Germany’s low insolation.
The same thing has happened in the US, where residential solar is said to be at grid parity in California. California’s average delivered residential electricity price (EIA March 2015) was $0.17/kWh. So EIA’s $0.13/kWh utility PV estimate ‘works’. Except that compares residential apples to utility oranges. So does Palo Alto’s rooftop PV LCOE estimate of $0.155/kWh before subsidies, linked above. But neighboring Arizona’s March 2015 residential rate was $0.115/kWh. In 2005, the year before California passed its 2020 renewables mandate, its average residential rate was $0.116/kWh—like Arizona today. PV is at California parity only because of electricity pricing distortions from its renewables mandate– an economically irrational self-fulfilling prophesy.
Thanks to Chile’s inherently high cost generation, Spain’s Abengoa just won an open bid auction to supply 950GWh of solar to Chile’s central grid over 15 years. Atacama Desert (highest insolation in the world) generation will come from two 110MW power tower systems and one 100MW PV. Since the auction was open to CCST and coal bidders, this might seem market evidence of unsubsidized CSP grid parity under those circumstances. Not according to Abengoa Solar’s own announcement about the award. Abengoa receives direct subsidies from Chile and the EU; Chile set a goal for 20% renewables by 2025. The facilities are partly funded by the Clean Technology Fund, a World Bank administered $5.3 billion clean tech aid program. Even in high cost high insolation northern Chile, CSP is apparently not at grid parity without subsidies.
Abengoa’s Solana parabolic trough CSP figured above is 70 miles southwest of Phoenix Arizona, where the residential rate is $0.115/kWh. It cost $2 billion for 250MW (net), of which $1.45 billion was covered by a federal loan guarantee. Plus Abengoa received ~$600 million in federal cash payments under the 2009 American Recovery and Renewal Act (ARRA2009) in lieu of the 30% Investment Tax Credit (ITC) on renewables, itself an optional alternative to the Production Tax Credit usual for wind. So Abengoa effectively has nothing invested. It has a contract to sell Solana’s utility generation to Arizona Public Service (APS) for 30 years at $0.14/kWh, which is 21% over the residential rate, and more than twice what equivalent dispatchable CCGT would have cost. APS made this unfavorable power purchase agreement (PPA) because Arizona mandated 15% of its power be renewable by 2025.
Brightsource’s Ivanpah power tower CSP figured above sits in the Mohave Desert on 3500 acres of federal land. It cost $2.2 billion for 377MW (net). It received a $1.6 billion federal loan guarantee, plus ~$660 million in federal cash payments in lieu of ITC under ARRA2009. It is also ‘free’ for its backers. But not free for California ratepayers of SoCalEd and PGE, who are purchasing its utility generation under a 25 year PPA at $0.185/kWh after time-of-day adjustments. That is 8% above California’s residential rate, 59% above what the California rate was before its 2006 renewables mandate, and 3x CCGT generation rates. Ivanpah ironically helps guarantee California residential PV ‘grid parity’, but only in a very costly artificial sense.
The fairly unique characteristics of each CSP installation make experience curve cost reductions unimportant. Since steam turbine efficiency is already mature, it is probably not possible for CSP to ever reach true grid parity even in high insolation areas. Abengoa and Brightsource might disagree; but then why are they lobbying to make the 30% ITC subsidy permanent?
As noted in a previous posting, all megawatts are not equal nor are they fungible. The portion of the bill associated with generation makes up roughly 50% to 70% of a typical bill. Conventional CCGT and coal base load costs about $0.06/kWh (see wind post). Intermittent generation that does not reliably reduce peak dispatchable generation cannot lower generation costs. It raises them by forcing flexing of daytime base load generation, which may reduce fuel efficiency and always reduces capital efficiency. Germany’s Energiewende illustrates the cost magnitude. E.ON just imposed a €4.5 billion impairment charge on its conventional generating assets because of renewables flexing. 
Solar is variable (day/night) and intermittent (clouds). How this affects grid backup cost depends on the grid location. Those ‘hidden’ costs are lower where peak is summer midday driven by air conditioning, such as southern California. The hidden PV backup costs are much higher if peak is winter evenings, such as the UK and Germany.
The benefits of residential solar are also inflated by a hidden ‘net metering’ subsidy. Net metered residences receive a credit for each kWh injected into the grid equaling the billed price of each kWh drawn from the grid. 44 US states including California have net metering. In California, utilities pay the customer ~$0.17/kWh for residential PV flowing onto the grid. The utilities could have purchased that electricity from conventional generators for about $0.06/kWh. The ~$0.11/kWh difference is a wealth transfer from non-solar to solar customers. Net metering means PV owners pay less, so non-PV customers have to pay more to cover total utility costs. California’s Public Utilities Commission (CPUC) estimates this will be $1.1 billion by 2020.
Such subsidies cannot be sustained as the proportion of residential PV on the system increases. The subsidies grow exponentially with increased participation as the subsidizing group gets smaller and the subsidized group gets larger. For example, if the annual subsidy cost non-PV customers only $1 at a 0.1% penetration level, it will grow to $10 at 1% penetration, $53 at 5% penetration, and $250 at 20% penetration. Germany’s Umlage illustrates this.
A practical solution to net metering’s hidden subsidy is suggested by Ontario. Because of the large number of summer vacation cottages little used in winter, Ontario electricity bills are split into a monthly delivery charge (just for being on grid, largely independent of kWh) and an electricity charge for kWh that pays for generation. But implementing that practice would reduce or eliminate ‘apparent’ residential PV parity.
PV is probably already at ‘true’ grid parity anywhere high insolation and inherently high cost conventional electricity generation coincide. This is, for example, the case in Hawaii where the average electricity price is presently $0.385/kWh, insolation is ‘yellowish’, and a number of utility PV installations have been built. It is credible that in reasonably insolated areas, PV will approach grid parity in the next few years even considering intermittency costs. The Southwestern US, Spain, southern Italy, northern Africa and its southern desert tip, India, and parts of Australia are plausible regions for eventual ‘true’ PV grid parity given their summer peak loads. The Saudi Arabian peninsula would be a candidate but for its abundant inexpensive natural gas and CCGT.
Renewables advocates can now show PV at ‘artificial’ grid parity where subsidies and/or renewable mandates have distorted electricity pricing, which is the case in Germany and California. PV parity becomes a lot easier if the price of electricity is increased 50 to 100 percent by fiat.
In the U.S., existing CSP projects would not have been built but for federal loan guarantees, the provisions of ARRA2009, and state mandates. All proposed but unbuilt U.S. CSP projects have been postponed due to expiration of most ARRA2009 provisions and uncertainty about the ITC set for reduction to 10% in 2016. These actions by CSP developers prove there is no US parity. So long as the US has abundant inexpensive natural gas for CCGT, there will never be.
Like wind, solar is ‘horses for courses’. Where conditions are favorable (California’s Mohave, Chile’s Atacama), some solar will sometimes make sense. Because of intermittency and ‘hidden’ grid backup costs, a lot of solar may not make sense even with summer peak air conditioning loads, as California is apparently in the process of learning. The pell-mell PV rush in the UK and Germany (where peak loads are winter evenings) only makes sense if CAGW is believed so dire a problem that high mitigation costs are of no consequence. We doubt that is the case, and suspect the UK and Germany will eventually discover reality the hard way as their conventional generators are driven out of business by renewables forced flexing.
 Forbes 11/10/2014 and 1/19/2015, FSLR 1Q2015 earnings call.
 First Solar PR 1/19/2015
 NREL/PR-6120-62558 (9/22/2014)
 Fraunhofer ISE, Recent Facts about Photovoltaics in Germany (1/7/2015)
 Abengoa Solar PR 1/9/2014
 FERC report for Ivanpah production and revenue January-September 2014