Monthly Archives: September 2013

Solar deployment – are there limits as costs come down?

A kWh is a different product depending on when and where it is delivered.  The rapid fall in the costs of solar PV implies that building grids, storage and commercial arrangements able to match supply and demand is much more urgent.  This will require strong policy drivers.

Falling costs are making solar PV increasingly competitive with other forms of electricity generation.  This post looks at what might limit solar PV’s deployment if costs continue to fall and reach levels low enough to allow for additional expenditure on grids, storage and demand-side infrastructure while remaining economically competitive.   I’m not taking a view on if or when this will happen, or how low costs might become – there is still a significant way to go to reach that point on a global basis.  I’m simply looking at what the remaining barriers would be if they did.  I’ll use some rough and ready numbers to look at what it might take for solar to produce around a third of the world’s electricity consumption.  I’ll assume this illustratively to be about 17,000TWh (out of a total of around 50,000TWh) by mid-century[1], which would be around 180 times the 2012 total of around 93TWh[2] of solar PV output.

As has often been noted, the solar resource is easily large enough to provide such large amounts of electricity.  Recent data from the US National Renewable Energy Laboratories (NREL) shows average US output of 70kWh/m2 based on total site area (i.e. not only the panels)[3].   Generating 17,000TWh on this basis (likely a conservative assumption, as panel efficiency is likely to increase over time) would require an area of around 240,000km2, less than 0.2% of the world’s land surface.  This is a huge area – about the size of the United Kingdom – but far less than the land devoted to agriculture, which uses solar energy to grow food.  And solar power can often make use of spaces – such as rooftops and deserts – that have few alternative uses.   Local planning and environmental concerns seem likely to become a more prominent issue as solar deployment grows.  However these concerns seem unlikely to place a fundamental limit on the industry globally.

A solar industry meeting a third of world electricity demand would be very large, but not infeasibly so.  It would require about 300GW of capacity to be added each year on average worldwide, around 10 times the 2012 installation rate, which has grown to its current level in just a few years.

However, matching the location and timing of supply to demand is a huge challenge.  Electricity at a different time and place is a different product and so part of a separate market.  Grids and storage help link these different markets[4] and so minimise load shedding (though some may still be required).

The first problem is geographical proximity.  In some cases (such as California and Mexico) demand is quite close to high quality solar resources.  However in densely populated countries with weak solar resources, such as the UK, the challenge is much greater[5].  Electricity may need to be brought from sunnier regions, especially in winter, requiring large scale transmission infrastructure.  There may be more local issues. For example in Japan, grid reinforcement will be needed to bring power from the north to more populous areas, crossing boundaries between regional utilities[6].  However problems in this respect may not be universal.  One recent study indicated that the German grid is already quite robust[7].

Matching the timing of output and demand is even more problematic.  Solar output is much peakier than system demand, and peak output and demand will often not coincide.  One indication is that load balancing becomes a significant problem when solar begins to account for more than around 10-15% of generation[8].

As solar penetration increases relative prices at different times of day are likely to shift, which may well cause demand to respond.  More sophisticated market arrangements and system operation are likely to become important features of most scenarios with extensive penetration of renewables.

Matching the timing of peaks by moving power from where the sun is shining to where the demand is located could imply  tens or hundreds of GW of power to be moved across continental distances.  This is because the point at which the sun is highest in the sky (around noon), when solar output tends to be at its maximum, moves quite quickly across the surface of the earth.  At the equator it travels at just over 1000miles/hour, implying that to service demand even an hour later in the day power must be moved hundreds of miles from west to east.  The chart below shows how far west you need to go to shift the time of peak one hour later at the latitudes of some of the world’s major cities.  To move the peak a quarter of the day – from a noon production peak to 6pm demand – you need to move power a quarter of the way round the world.  And the direction does not always help.  To meet later demand on the US west coast solar panels would need to be out in the Pacific Ocean rather than Arizona and Texas. ( Putting more west facing panels in California itself helps this.  There is some loss of total output but the match to system peak improves).  China, with its population concentrated on the east coast, is better served in the evening, but would run into problems in the morning.  This implies that load balancing using transmission will be a huge challenge from a technological, regulatory and commercial perspective.

Solar production needs to be hundreds of miles west to meet a demand peak one hour later in the day at the latitude of the world’s major cities…

Chart of distance

Even the most extensive links may not be enough on occasions when the sun is over the oceans.  The map below shows where the sun is shining at midnight GMT on 21st December.  There is an hour or so of setting winter sun still left on the US west coast, and the weak first hour or two of the day’s output from panels in east Asia, with Australia in daylight (and therefore with some intriguing export possibilities if links can be built far enough).  But the whole of Europe, Africa, the Middle East, India, and almost all of Russia and North and South America and a good deal of the rest of Asia are in darkness.

daylight map

Building storage to address this problem is challenging because of the huge scale needed, as well as because of  the cost.  The subject is too large to go into here in detail, although one recent study showed it to be crucial for reaching a third of supply in  North America  [9]. In northern Europe very large amounts of storage are required even to balance load within the day. Seasonal storage (because for example average intensity of sunlight in the UK is nine times higher in summer than in winter) would require enormous capacity[10]Germany’s subsidy for storage as part of new residential PV systems, which was introduced in May, and California’s plan for 1300MW of storage by 2020 are early examples of the type of initiative that is likely to be required.  Among other effects the premium for hydro power for load balancing is likely to increase.  And reductions in load factor due to no storage being available and so surplus remaining unused at peak will be less of a problem the lower the capital costs of solar become.

Building transmission and storage infrastructure, along with the arrangements to manage them, will take decades at the scale required.  And getting the costs of storage down will also be hugely challenging.  This will be accompanied by the need to make significant changes to market mechanisms so that they can more effectively balance supply and demand.  None of this will be achieved easily, and strong policy drivers are likely to be required for this to happen as fast as now looks likely to be required if solar is to play a central role in decarbonising power systems.

Few expected solar to become quite so cost competitive quite so quickly.  This largely unanticipated increase in competitiveness leads to a similarly accelerated programme now being required to build grids and storage able to incorporate increasingly large amounts of solar into the world’s power systems.

Adam Whitmore  –  25th September 2013

[1] This is broadly similar to Shell’s Oceans Scenario, which shows 20,800 TWh of solar generation, 36% of a total of 57,800TWh.  The total consumption considered here is based on an extrapolation to 2050 of the IEA’s New Policies scenario for 2035 to 2050.  This may be higher with increased electrification of end use.  It may be lower with greater efficiency, but in any case only intended to indicate order of magnitude.  See

[2] Source BP Statistical Review of world energy, 2013

[3]  I’ve taken the average for large solar of 3.4 acres per GWh.

[4] Consumers cannot readily substitute between consumption in different places and at different times – you need electricity in your living room now, and electricity in someone else’s living room later is not the same product.  A hypothetical monopolist could profitably impose a small but significant non-transitory increase in price, implying that the markets are separate.

[5] David MacKay. Solar energy in the context of energy use, energy transportation ans energy storage.  Philosophical Transactions of the Royal Society Vol 371,number 1996.

[10] See reference 3 above for a discussion of this point.

The social cost of carbon as a marker for price floors

Estimates of the social cost of carbon can help set floor prices in emissions trading schemes.  Estimates are likely to suggest a floor price for the EUETS of around €20-30/tCO2, rising over time. 

(Note: for a discussion on the social cost of carbon itself see the page on this issue under carbon pricing).


Those arguing against any price management in the EUETS sometimes suggest that, first, there is nothing wrong with the currently low level of prices, as it simply means that targets have been met at low cost, and, second, that there is no objective basis for setting the level of a floor price.

However neither of these propositions is valid.  We can be fairly sure that current prices in the EUETS are too low because prices are below the bottom end of the range of estimates for the damage caused by emissions – the social cost of carbon (SCC) – and so violate the principle that the externality of environmental damage should be fully included in the marginal price of emissions.  The emissions reduction target to 2020, an inevitable trade-off between costs and benefits, was, with hindsight, insufficiently ambitious[1].  And the SCC provides an objective, non-arbitrary method for setting a price floor, although inevitably the quantification is subject to uncertainties.

Estimating the SCC involves many uncertainties about the nature, extent and cost of damages, and the value that should be placed on non-market effects, including many changes to ecosystems.  These uncertainties are compounded by the damage being determined by the stock not the flow of GHGs, and even by the cumulative effect of the stock over time, with effects depending on how long a given concentration of GHGs is present in the atmosphere.  Furthermore the effects of changing the stock are likely to be highly non-linear effects.  There is also considerable uncertainty about the discount rate that should be used in weighting current and future costs, and the adjustments necessary to account for effects across people at very different levels of income (equity weighting).  These uncertainties lead to a wide range estimates of the SCC.

The US Environmental Protection Agency (EPA) has recently published new estimates of the SCC.  These are intended for use is assessing the cost effectiveness of policies such as fuel efficiency standards for vehicles.  Because many damages are large but occur in the distant future the choice of discount rate has a particularly large effect on the results, so the results are presented for a range of discount rates.  Values in $2011 range from $12 to $117/tCO2 in 2015 (around 2013€9.50- €92/tCO2) rising over time by roughly 2% p.a. in real terms.  This range, covering an order of magnitude, is fairly typical of surveys of the SCC, though some have argued for much higher values, and a few for lower values[2].

Social Cost of CO2, 2015-2050 a (in 2011 Dollars)

Discount Rate and Statistic


5% Average

3% Average

2.5% Average

3% 95thpercentile









































a The SCC values are dollar-year and emissions-year specific.  Source: USA Environmental Protection Agency[3]

There are some who have argued that the uncertainties are so great that such analysis is all but useless for policy making[4].  However this seems to go too far.  Although wide, this range is a useful guide for decision making because it gives guidance for appropriate minimum levels for the carbon price, demonstrating clearly that the current EUETS price is almost certainly below the cost of the damage caused by emissions of greenhouse gases.  These estimates can thus act as useful markers for the level of floor price that should prevail under the EUETS.  Indeed at a 3% discount rate the SCC estimate for 2020 of $46/tCO2 is quite close to the target UK carbon price floor of £30/tCO2.

Many of the limitations of the modelling suggest that there are good reasons to suppose that any floor price should be higher than the bottom end of the range of estimates[5].

First, such a low estimate follows from the application of a constant, continuously compounded discount rate of 5%.  It is likely to be more appropriate to use a discount rate that falls over time, which would reduce the effect of discounting and so increase the estimated SCC towards the higher values in the table.  A workshop held by the EPA acknowledges this as a consensus view, and that this approach is the adopted by the UK and France[6].

Second, the lower estimates of the SCC also exclude some effects, and fail to take adequate account of some high impact events (upper tail of the distribution).  Even the higher estimates will often exclude material damages [7].  Inclusion of these would lead to higher damage estimates.  Indeed, the discontinuities raise fundamental challenges to the concept of marginal damage from a tonne of emissions.  Furthermore, even for those damages that are included models used to estimate the costs of damages may not yet include the latest evidence on impacts, which in many cases suggests higher levels of damage.  Underestimates of damages are likely to be especially large at high temperature changes.

On balance it thus seems likely that an optimal floor price is above the bottom of the range shown by the EPA, perhaps around the value for a 3% discount rate of around $40/tCO2 (€30/tCO2), although a case for higher numbers can readily be made.  However corresponding EU analysis would be likely to be required in practice to underlie the setting of any floor price level, and pragmatic considerations may lead to the choice of a value below this, perhaps in the range €20-30/tCO2.  It may be desirable to converge to this level over a few years in practice to prevent significant price discontinuities.  This would of course need to be accompanied by adequate shielding for sectors at risk of carbon leakage.

If the price of EUAs were to rise towards the top of the range of estimates for the SCC then it may imply that some abatement would be more expensive than the damage caused by the emissions, and so would not be warranted on grounds of economic efficiency.  This may suggest the need for a price ceiling at around this level.  However, the case for this is much less clear-cut than for the lower bound, due to the possible presence of the discontinuities, high impact events, and non-market costs already alluded to.  This may imply that only a higher ceiling price, or even no ceiling price at all, is appropriate if this takes the form of unlimited additional allowances being made available (fuller discussion of this point, including how it relates to abatement efforts elsewhere as well as the political economy dimension, will need to await another post).  However the upper end of the range of SCC estimates may at least form a marker for an appropriate price at which to release allowances held in a price containment reserve of allowances taken from within the cap, such as that found in California, providing a “soft” ceiling until the reserve is exhausted.

The California and Quebec schemes have floor prices approximately consistent with the range of SCCs, with price floors at $10/ tCO2.  Although a little below the estimates shown in the table this is indexed at 5% real p.a., and so will quickly converge with the levels shown if this indexation is retained.  However RGGI has a price floor (around $2/short ton) that looks much too low by this metric, with a strong case for a substantial increase.

Estimates of the SCC may be uncertain, but they nevertheless represent a useful way of putting at least a lower bound on the price in an ETS.  The use of uncertain estimates is better than allowing the price to fall to close to zero, a price which is surely wrong.  And it is likely that estimates of the SCC in a European context would imply a floor prices well above current EUETS prices.

Adam Whitmore – 11th September 2013 (with a minor update 2nd April 2014)

[1] The marginal price signal is at too low a level, so some economically efficient abatement is not being signalled.  It is possible that an inefficient mix of abatement is being purchased, even though the level of abatement is efficient.  This could be the case if, for example, there was too much expensive abatement through renewables programmes.  In this case the appropriate response would be to reduce the expensive abatement under other programmes, in which case the price floor in the ETS would still be appropriate but might not bind.  Alternatively, renewables programmes may recognise the presence of other market failures, notably those associated with failure to recognise spill-over effects from innovation, and the policies in place are appropriate.  In this case the cap is too loose.  Some mix of these explanations is of course possible, although the latter seems the more plausible.  In either case a floor to maintain an efficient marginal price signal remains appropriate.

[2] Comparing estimates of the SCC from different sources presents a number of difficulties, including the currency and year in which the estimates are quoted (US$ 2011 here), the date of the estimates and the assumptions made, all of which are sometimes unclear.  However, most estimates seem to imply values roughly within the US EPA range.  For a discussion of a variety of issues see:  Also see: L. Johnson and C. Hope, “The social cost of carbon in U.S. regulatory impact analyses: an introduction and critique” Journal of Environmental Studies and Sciences.  September 2012, Volume 2, Issue 3, pp 205-221.  More recent work by C. Hope and M. Hope (“The Social cost of CO2 in a low growth world” Nature Climate Change, August 2013) points out that with lower growth estimates of the SCC rise as future generations are correspondingly poorer (essentially this is an aspect of the discount rate issue).  A few calculations by others in the past have produced very low and even occasionally negative values for SCC (implying GHG emissions are beneficial), due, for example, to increased agricultural productivity, but these do not seem a plausible reflection of current circumstances and understanding.

[4] See  The author argues that the assumptions, especially on the damage function, cannot be sufficiently robust to base conclusions on.  However he acknowledges the pragmatic value of such results as a possible marker for a carbon price.

[5] For estimates giving higher values see:, which explicitly critiques the US EPA analysis.


[7]  A good survey of omissions from calculations of the SCC is given by a recent report co-sponsored by the US NGOs the Environmental Defense Fund and National Resources Defence Council