Geographical wind smoothing, supergrids and energy storage

Guest Post by Jani-Petri Martikainen. Jani-Petri is a theoretical physicist doing fundamental research in the field of ultracold quantum gases. Most of his current research activities are computational and involve bosonic or fermionic atoms in optical lattices. He has a lively interest on environmental, climate, and energy issues. He runs the blog PassiiviIdentiteetti, which is mostly written in Finnish.

————

By aggregating power generation from wind farms spread across the whole (North Sea) area, periods of very low or very high power flows would be reduced to a negligible amount. A dip in wind power generation in one area would balanced by higher production in another area. European renewable energy council and Greenpeace (page 34).

Strangely, proponents feel comfortable in making such statements, but show a noticeable lack of interest in actually demonstrating whether the statements are true. Why is this? In science the burden of proof falls upon the claimant and it would be desirable if the same  principle were to apply to discussions about energy policies. (Notice by the way, that EREC+GP are not even satisfied with claiming that wind speeds in different parts of the North Sea are uncorrelated, but actually claim that speeds are anti-correlated.)  Why is it, that an amateur like me [in energy analysis] feels the need to do his own computations to figure out such issues rather than just being able to read proper studies online?

Since it appears difficult (certainly outside academic journals) to find detailed numbers on how strongly, for example, wind power actually relies on fossil fuels, I decided to do some estimates myself. I am not primarily interested in cosmetic amounts of wind power production, but will take the ambitious renewable visions seriously and study scenarios where wind power would be enough to power our entire society. I want to understand to what extent electricity production in such scenarios still relies on reliable energy sources and what kind of energy storage is required to enable wind power to stand on its own feet. Since hydropower capacity at a global level is limited, I will mostly use the term “reliable energy source” as an euphemism for fossil fuels. Not to be too parochial and allow for massively distributed generation,  I will assume a “super(duper?)grid” coupling wind power sources from three different continents together.

As a starting point I want to create a production profile based on real wind power production data. As sources I choose south-eastern Australia, Ireland, and the Bonneville Power Administration in Oregon, US. Each has roughly comparable amounts of wind power installed, but I will scale the capacity of each to 3333 MWe so that the combined capacity will end up being 10 GWe (peak). Data for BPA and Australia is given every 5 minutes while the Irish data is every 15 min.

To get the datasets to match I will make a linear interpolation of the Irish data. Furthermore, since my chosen time period for the Australian data (1.8.2010 — 30.7.2011) is a bit different from the other two (1.7.2010 — 30.6.2011), I will fold the Australian data onto itself from the end to generate few missing datapoints. I take the consumption profile from the BPA load, but reserve the right to change its scale to suit my purposes. As a result, I get a combined wind power production from three massive clusters of wind turbines on three different continents. (Note: A slight bias might be caused by increasing capacity over the year.) In Figure 1, I show the power distribution for the individual clusters and for the combined system. The distributions look a bit different to each other, presumably because the Australian turbines are the most distributed geographically. The combined system has about a 7% probability of producing less than 10% of the installed capacity.

Figures 2 and 3 show how production and consumption relate to one another during one randomly picked week in two different scenarios. In the Figure 2, the minimum consumption is the same as the maximum production, so that no wind power has to be wasted. In the Figure 3, the wind power produces the same amount of electricity in a year as the society consumes. Because in neither case does the consumption match the production, some reliable source of energy must bridge the difference. For now I assume that this reliable source of energy can be turned on instantaneously in response to changes in wind production. This assumption is typically wrong and eventually I will make few remarks as to how serious this assumption is.

For the scenario in Fig. 2 it turns out that 74% of the electricity is produced with fossil fuels and the capacity of the reliable power plants must be 92% of the peak demand. The CO2 reductions in this scenario are nowhere close to what is required and the entire wind capacity has been build to work in tandem with power plants burning fossil fuels. This modest tinkering of electricity supply is quite close to what is being practiced today in many countries.

In the ambitious scenario presented in Figure 3, some of the wind power ends up wasted and periods of low production must be covered with fossil fuels. It turns out that the capacity factor of wind power drops from around 30% to about 24%. Power plants burning fossil fuels cover about 21% of demand and their installed (standby) capacity must be 88% of peak demand. If we take the threat of climate change seriously, even this rate of emissions is excessive, given that electricity production is not the only source of greenhouse gases and that the global electricity consumption will most likely rise. Importantly, it should also be noted that in this scenario the reliable power plants are running at a capacity factor of only 15%, which increases the cost of their power dramatically. Under this scenario one would quite likely (and perversely) end up paying subsidies to the owners of the power plants burning fossil fuels.

(As an aside,  a leaked European Commission document apparently includes a 50% wind scenario by 2050. Based on the above approach this would imply a need for reliable power plants that can account for 92% of the peak demand. Capacity factor for these standby plants would be around 35%. Since solar PV production almost never peaks during the peak demand period and is reliably off during most of the day… CSP with storage might be theoretical possibility, but to be able to contribute to next days peak demand and compensate for the cloudy days they will need substantial storage. Solutions where CSP plants are backed up with fossil fuels are clearly not satisfactory.)

That these scenarios rely fundamentally on fossil fuels does not feel right to someone seriously concerned about climate change. This dependency can be broken if wind power during periods of high production could be stored somewhere. How much storage would be needed? I will now assume that: (i) (only) 20% of the energy is lost during the transfer of wind-generated energy to and from the storage, (ii) storage doesn’t “leak”, (iii) there are no limits on the storage input-output powers, and (iv) that the storage is arbitrarily large. Only type of storage that might approach these conditions even to some extent, appears to be pumped-hydro storage.

In Figure 4 I show how the energy content in the storage varies over the year. I choose the consumption to such a level that the storage at the end of the year is about the same as in the beginning of the year. It turns out that the entire electricity consumption (95% of wind production) could be covered with wind power if the storage amounts to about 9% of the yearly production, or 2.5 million MWh. In practice, about this amount of energy equates to that which would be released when the water from a 90km² lake that is 20 meters deep drops 500 meters. Naturally this same volume of (fresh) water would also have to be stored at lower elevation to await pumping back into the mountains. However, this scenario appears somewhat unrealistic in that it requires that we can store energy at a power 5.1 GWe and release it at 4.3 GWe. These figures are massive relative to the maximum demand of 4.7 GWe.

So let us proceed to make things perhaps a bit more realistic by throttling the storage input-output power to “just” 1 GWe. In this case some of the wind power is again lost and dependence on fossil fuels reappears. Consumption can now be 89% of yearly wind power production and storage must be sufficient for about 5% of production. In this scenario, 4.5% of consumption would be covered by reliable power plants running with a capacity factor of just 5%. However, their installed capacity must still be 63% of peak demand. If we throttle the storage power further, the need for fossils fuels increases.

What if we just store energy for few days? If the storage is for 5 days peak production, and we throttled its power like before, about 9% of consumption must be covered with reliables. Their installed capacity must be 70% of peak demand and the capacity factor is 8%. If we are to remove reliables entirely from the picture, the consumption must drop drastically to the average level of about 900 MWe. Naturally, this implies a drop in the winds capacity factor to less than 9%.

So far I have assumed that reliables can react instantaneously to changes in wind power production. Let us add a delay of 10 — 30 minutes to the scenario of Fig. 3, where most electricity was from wind. i.e., I assume that if the reliable source was turned off, it takes 10 — 30 minutes for it to start producing power again. In Figure 5 I show the resulting difference between production and demand. As is clear from the diagram, even with only 10 minutes delay, more than 600 MWe of mismatch can appear. Smaller discrepancies appear regularly over the year and their frequency increases as the reliables response becomes more sluggish. These observations presumably set some constraints on the amount of reliable power plants which must either be constantly spinning no matter what the wind conditions are, or be able to react very rapidly to changing wind conditions (hydro probably).

It is of interest to check what we actually gain by (hypothetically) combining the time-matched production from Irish, Australian, and BPA productions within a “super grid”. If we only use the production data from Australia for the wind-dominated scenario, 24% of the electricity would come from reliables (c.f. 21% with the “supergrid”) , required reliable capacity would be 92% (vs. 88%), and the reliables capacity factor would be 17% (vs. 15%). Therefore, it seems that distributing wind turbines over an area larger than around 1 million square kilometers (on three continents) provides only modest additional benefits. These benefits should naturally be balanced against the additional costs.

In all the above I have taken the consumption pattern to be fixed. In  principle, using smart grids the consumption could change and become more flexible. However, not only does the required change have to be very rapid, but it also has to be potentially a very large fraction of the total demand at certain times. It is naturally partly a political and ideological question as to whether it is desirable to force society to adapt to failures of the chosen technology, rather than demanding that the technology adapts to way people behave. (The way I phrased it, makes it quite clear where I stand.) In fact it is curious how eagerly proponents of, for example, wind power, wish to rely on smart grids even though the most obvious use of smart grids seems to be almost diametrically opposed to their vision! Let me explain.

Sending more detailed pricing signals to consumers, has the potential advantage of lowering peak demand and perhaps inversely increasing night time demand. Under such circumstances, the difference between average and peak demand is reduced and the share of the baseload power actually increases. In the extreme limit case, we would end up with an electricity supply entirely made out of baseload power plants (coal or nuclear typically). Not only would this lower the cost of average kWh, but it would also seem to simplify the design and maintainance of the energy infrastructure. i.e., used in this way, a smart grid seems to be a really smart idea! However, the way proponents of unreliables intend to use smart grids is quite different. For them, the smart grid is a way to lower demand not when demand is necessarily high, but when their favored energy supply is failing. The smart grid is then transformed into a system of managing (avoiding) brownouts, blackouts and load shedding. Used in this way, it is about giving consumers the choice between very costly energy and a blackout. A managed blackout is certainly better than unmanaged one, but how is it exactly better than not having blackouts is unclear.

For the reasons above, I think it is clear that it is very difficult to base the electricity supply on erratic sources of energy. As soon as we start estimating the required storage capacity or the capacity of reliable backup power, we end up with massive figures, implying huge escalations in costs, or an unacceptable reliance on fossil fuels. Getting the production and demand to match each other becomes ever more complicated, and I cannot help myself thinking that the resulting device starts to, more and more, resemble a Rube Goldberg device.

Here is the scenario you end up with: If wind power is too variable, build a supergrid and then connect it to smart grid. Then combine everything to solar power on the other side of continent, which will be smoothed with wave power coupled to geothermal, and as an icing on the cake build large number of microlevel bio, natural gas,and hydro power plants across the continent. If the goal is to create a reliable voltage difference, is there no easier way? If the goal is to redistribute common resources to those who manufacture pieces of the device, however,  then there is unarguably some internal logic. Also, if the real goal is to maintain de facto dependence on fossil fuels, this approach is eminently sensible.

Unfortunately, this kind of confusion makes it harder to understand how the whole system works (or doesn’t work) and also harder to understand the final costs and emission levels involved. (Sometimes I get the feeling that proponents for unreliables prefer it that way.) Such visions do not become more convincing when one observes the politics involved. Each part of the device is constructed with scant regard for other parts, with multitude of different national (especially in the European Union) subsidy schemes. Many parts of the device also seem to be more like rhetorical tools, to divert attention from the shortcomings of the activity under spotlight. For example, supergrids are often evoked as a tool that would make huge fluctuations of wind power at national level disappear. As we have seen from this analysis, not only is this assumption unjustified, but it also seems unclear who exactly is supposed to pay for such grids. It is certainly not included in the typical cost estimates for wind power. Also, how are the small nations around Germany, some of which have no need for wind power, supposed to balance the wind power of 80 million Germans? Would Germany pay for the construction of wind turbines in, and transmission cables from, another country? Not likely.

Only scenarios which are based on reliable energy sources from the beginning seem to avoid the problems discussed here. Scenarios based on unreliable sources become progressively harder as their share of electricity supply increase. Reducing GHG emissions sufficiently requires, in practice, total decarbonization of the electricity supply, and the emissions reductions achieved by the time erratic sources run into trouble are far too low. I cannot avoid the conclusion that approaches based on renewables will mainly, at a very large expense, end up delaying the real decisions we must eventually make to lower emissions to acceptable levels. The alternative zero-carbon baseload source seems rather obvious…