Sunday Train: Portfolio Theory vs the Myth of Intermittent Wind Power

(2 pm. – promoted by ek hornbeck)

This last week, in the comment section of the EnergyCollective, I saw the same myth that I have seen time and time again regarding wind power:

Fact 1: renewables are aleatorically intermittent, and so unreliable.

Fact 2: due to Fact 1, they cannot provide energy when it is needed, but only when and in the quantity they can

Fact 3: users have to get energy when they need it, not when it is aleatorically provided

Fact 4: to date, there is no storage system that can be useful for a complex industrial society

Fact 5: due to facts 1 to 4, renewables need to have a back up system that can cope with the needs of the users.

Fact 6: that back up system cannot be just stopped and then put to generation in a few seconds or minutes, and usually have to generate at low efficiency to maintain the back up at call point, generating added costs, besides the usuals as maintainance, lost profits, complex distribution grid, etc.

… not surprisingly ending with climate crisis denialism in “Fact” 8, since the name of the game here is clearly not arguing by starting with facts and seeing what conclusion you arrive it, but rather is myth creation and propagation in support of an already selected conclusion.

While many people don’t know what “aleatorically” means, many would actually share the misconception that windpower is an intrinsically intermittent resource. However, for wind power, the “Fact 1” is in many cases “Falsehood 1”. Even though individual wind turbines are intermittent, for many wind resource regions, it turns out that a substantial share of wind power is not intermittent at all, in either their “by chance (aleatorically) and unpredictable” component or their “by chance (aleatorically), though predictable” component.

Portfolio Theory

Lets consider the western part of the United States, and four of the main electrical regions of the West: the regions of the Midwest Independent System Operator (MISO), Bonneville Power Authority (BPA), the Electricity Reliability Council of Texas (ERCOT), and the California Independent System Operator (CAISO). The map is from Fertig, et al (2012, incl. Open Access), and includes locations of existing windfarms in these four regions.

Now, each and every one of those windfarms are intermittent and volatile energy sources, in that there is some part of each year that each and every wind farm is generating no power. For each individual wind turbine, a substantial part of the volatility of the energy supply is predictable a day or more in advance, while a substantial part changes from hour to hour, and there is a component of the electricity supply of each individual wind turbine that can change from minute to minute and even second to second.

However, that does not mean that when you combine a number of wind turbine in the same wind farm, a number of wind farms from a single electricity supply region, or a number of wind resources from distinct electricity supply regions that the generation of that portfolio of wind turbines has the same intermittence and volatility. As described in H. Rao (2012, a Masters Thesis):

According to literature already published, the idea of portfolio diversification has been considered to reduce the variability in wind power output due to the reduction in the correlation between wind patterns. Geographic locations that are spread far apart are able to diversify the ‘risk’ of the entire wind power resource because while some sites encounter poor winds, there exists a possibility of another site encountering a state of high winds, thereby compensating them. Thus wind farm geographic diversification can be employed to smooth out the fluctuation in wind power output from uncorrelated sites. This is currently achieved through wind power variations in one part of the country canceling out variations in wind power in another part of the country as claimed by Drake and Hubacek [1], depending on the distance between them. The reduction in variability is related to the distance between sites, which stems from the reduction in correlation as distance between sites increases. For instance, Sinden found that the hourly correlation coefficient between U.K. wind farm sites decreases to approximately 0.1 over distances in excess of 100 km [12].

Fertig et al (2012, linked above) find that the high frequency volatility of power is substantially reduced at the wind farm level, with the interconnection of four windfarms reducing the ratio of very high frequency variability to hour by hour variability by 87%. This is one of the reasons that utility scale wind turbines are normally installed in the groups that we call a wind farm rather than as single wind turbines, since the lower variability of energy supply at the second, minute and quarter hour time scales of a group of wind turbines makes for substantially easier integration into the regional power supply.

All that is about variability, but what about intermittence? In the “duration curve” plot to the right, from Fertig et al. (2012), the figure plots what percentage of total capacity is available for what percentage of the hours during 2009, based on production of the windfarms noted on the above map.

The first thing that is immediately clear is that if all four regions are interconnected, there was never an hour in 2009 in which there was no windpower being produced somewhere. Drilling down into the individual regions, they give availability for the “firm power” range of 79% to 92% (they do not cite CAISO separately):

• All four regions: 17%-12% of capacity available 72%-92% of the time
  
• MISO: 13%-6% of capacity available 72%-92% of the time
  
• ERCOT: 10%-4% of capacity available 72%-92% of the time
  
• BPA: 2%-0.2% of capacity available 72%-92% of the time
  

These figures are deceptive, however, because the gross installed capacity is a deceptive measure of windpower, as discussed in more detail by by Robert Wilson at the Energy Collective. Because wind power (and solar power) is variable, the energy available from the sources over the course of a year are substantially less than half of the headline capacity. For example, based on a rough linear interpolation from the duration curve of the four regions, the average electricity supplied by windpower in these four regions was about 27% of headline capacity.

So we need to adjust these figure so that it describes firm power as a share of total energy available from windpower. For simplicity, as well as to reflect the fact that newly installed wind turbines have higher capacity availability than wind turbines installed ten or twenty years ago, I will use a 30% capacity factor. That gives the share of firm power as:

• All four regions: 56%-40% of windpower supply 72%-92% of the time
  
• MISO: 43%-20% of windpower supply 72%-92% of the time
  
• ERCOT: 33%-13% of windpower supply 72%-92% of the time
  
• BPA: 7%-0.7% of windpower supply 72%-92% of the time
  

Now, depending on how much on-demand firming capacity is available, from one fifth to two fifths of windpower in the Midwest ISO region is variable, but not intermittent, and if all four of regions were connected by cross-haul transmission, roughly half of the windpower supply is firm power, and about a fifth of the total energy supply is always available.

One implication of these figures is, as Fertig, et al. (2012) notes, “BPA is the region that would benefit most from interconnection with other regions”, which is reinforced by fact that the correlation of BPA with its eastern neighbor, MISO, is (-0.06). Indeed, the BPA could well benefit by acting as a bridge between MISO and CAISO, with the negative correlation between MISO and CAISO being an even more substantial (-0.26).

 

Firming Windpower with Hydropower

If you fall into the myth-making “Facts” listed at the beginning of this post, firming windpower seems like an almost insurmountable challenge: “back up system cannot be just stopped and then put to generation in a few seconds or minutes” is premised on the idea that much of the challenge lies in bringing back-up systems on line in seconds or minutes. Digging into Fertig, et al. (2012: p. 3) debunks this for any well-interconnected system:

… power fluctuations at frequencies corresponding to 10 min, for example, are at least a factor of a thousand smaller than those at periods of 12 h. This property has important practical consequences: if the PSD of wind were flat (white noise), large amounts

of very fast-ramping sources would be required to buffer the fluctuations of wind power. The negative slope of the PSD implies that slow-ramping resources such as coal or combined-cycle gas plants can compensate for most of wind power’s variability, with less reliance on fast-ramping resources such as batteries and peaker gas plants.

So:

… Balancing wind with a portfolio containing fast-ramping resources such as batteries, fuel cells and supercapacitors, in addition to slower-ramping resources, would avoid the unnecessary expense incurred by building a single type of linear ramp rate generator that would have excess capacity at low frequencies.

Since the required capacity of the very fast-ramping resources is less than 0.1% of the required capacity of the slower ramping resources, the primary firming challenge is at the scale of hours and days, not at the scale of seconds and minutes, no matter what is implies by the scare-mongering language of the above Mythical Fact number 6.

Back in September, 2013, I have discussed one slow-ramping carbon-neutral electricity supply, biocoal produced from sustainable coppice or perennial grasses. However, it should be kept in mind that being able to use a slower ramping supply means that it can use either slower or faster ramping supplies.

In this region, one of the firming power resources already available is the substantial installed hydropower generating capacity of the Columbia River and its tributaries. Hydropower has exactly the qualities we look for in firming power sources, which is that power that it can bring substantially more power online at a given point in time than it can maintain over the course of a year.

Why is this what we are looking for? Consider the challenge of firming 1,000MW of headline capacity windpower from various wind farms spread across the MISO region. First, that is not 1,000MW of supply on average over each hour of the year, but something closer to 300MW of supply. If we wish to have 40% of that available as firm power on an ongoing basis, that means that we need up to 90MW of capacity, for the rare times when there is very little windpower being produced in the MISO region. However, 79% of the time, that is entirely provided by windpower, and 92% of the time half of it is provided by windpower, so on a crude linear approximation, while 90MW generating capacity has to be available at any given point in time for firming, only 8.4MW of hydropower is used on average for firming.

To put this into scale, consider that the Grand Coulee hydropower dam in the State of Washington in the BPA region has a capacity of 6.9 gigawatts (6,900 MW), so 3GW, less than half of its capacity, would be sufficient to firm 40% of the electricity supply of MISO hydropower generating 7.5GW on average, which would amount to 25GW of installed capacity. To put that in scale, that is about 25,000 GW-hrs/yr of firm electricity supply and 65,000 GW-hrs/yr of total electricity supply, for a state that presently produces about 100,000 GW-hrs/yr.

A step beyond existing hydropower capacity is pumped hydropower. Pumped hydropower (PHEV) involves consuming power when it is relatively abundant, and therefore less expensive, to pump water from a lower reservoir or source to an upper reservoir, and then using it to generate electricity by allowing it to return to the lower reservoir or source. JG Levine (2007, Master Thesis. pdf) considered nine sites in Colorado that might be suitable for PHEV development, with generating capacity of 3GW and energy capacity of 21GWh/yr.

The economics of PHEV is straightforward. If the storage energy efficiency of the system is 80%, the cost of power used to fill the upper reservoir is $0.03/kWh, and the price paid for power when it is generated is $0.13/kWh, then the return on the power stored is $0.08/kWh. Estimating the total capacity of the PHEV site and the cost of construction, Levine finds that the period for the sites he studies to pay back their original construction cost ranges from 15-35yrs.

And, of course, those are payback periods under status quo assumptions. If we were to impose adequate carbon charges, one impact would be an increase in the wholesale price of peak power provided by natural gas peaking plants, which would increase the price paid to the PHEV site delivering stored power. And it may be possible to contract to acquire surplus windpower, when available in the system, at a discount rate, which would reduce the average cost of power used to fill the upper reservoir. Either or both factors would accelerate the payback period on PHEV systems.

Most existing PHEV facilities in the US are part of conventional dammed hydropower systems, with either the upper reservoir, lower reservoir, or both providing by damming a river. According to Yang and Jackson (2011), after a long period with little interest in new PHEV systems, the US has seen a recent increase in interest, with FERC in 2010 listing 32 PHEV projects receiving preliminary permits and four more with applications in process. It is notable that fewer than a quarter involved a new dam for either upper or lower reservoir, a majority of both upper and lower reservoirs are not on an existing body of water at all, several projects use an abandoned quarry for the upper reservoir, and several projects feature a lower reservoir that is underground

 

Climate Suicide Club Myths Posing as Facts

Between existing conventional hydropower, opportunities for new PHEV development, and carbon-neutral renewable power sources such as Biocoal that can be produced on schedule when required to provide 12hr firming, where does that leave Myth #4, that there is no storage system that can be used by used for a complex industrial society?

It would seem to leave it in tatters. The myth rests on exaggerating the intermittency of windpower, our most abundant, highest EROI variable renewable energy source, the volatility, ignoring the feasibility of new slow ramping dispatchable renewable energy sources like Biocoal and the ability to use the existing renewable energy resource of hydropower to firm windpower, all of which would leave us exaggerating the stored energy requirement that goes with substantial reliance on wind and solar power. It then requires us to directly ignore the already existing energy storage technology of PHEV.

I do want to touch on a final point from Fertis, et al. (2012). They compare transmission as a firming technology to provision of natural gas peaker plants as a firming technology by examining what length of transmission corridor between the two regions with the greatest wind variability, BPA in the Pacific Northwest and ERCOT in Texas. The cost of sufficient gas turbine capacity to provide the same firming as the 1400 mile corridor would cover 490-740 miles of transmission capacity, or 630-960 miles if emissions damages are included.

While it should be noted that this is a fairly rough first-order estimate, this is an interesting exercise. However, at the broadest level, it demonstrates the difference between value-based emissions damages and quantity-based emissions damages. In the case of CO2 emissions, a notional value of CO2 emissions at $50/ton is not actually saying that CO2 has $50 worth of environmental cost per ton, so that only emission reductions that cost less than $50/ton should be attempted. Instead, given the fact that we must refrain from emitting a majority of the carbon in fossil fuel reserves in the ground, it means that $50/ton is sufficient to keep the carbon from being emitted. If it turns out that $50/ton is not sufficient to keep the carbon from being emitted, then CO2 has to be priced higher than $50/ton.

On a more detailed level, this analysis is a reminder that when we are considering regional inter-connects, we should not lose sight of the resource geography. Given the BPA’s substantial hydropower resource and the substantial mismatch between windpower capacity and in-state power consumption in the western side of the MISO region, surely the strategic transmission corridor is between the BPA and the MISO.

This is amplified when we consider strategies for reducing the incremental cost of adding substantial inter-region transmission capacity. As I have discussed on a number of previous occasions, and as George Berka noted last Fall in Inside EV’s, a Steel Interstate system of long-hail electric freight corridor is an essential element of a system of sustainable freight transport in this country.

At the same time, however, establishing a Steel Interstate corridor offers the opportunity for cost sharing with long-distance HVDC transmission corridors, by establishing the HVDC corridor along the Steel Interstate itself. For example, the BNSF Northern Transcontinental corridor (which the Amtrak Empire Builder runs on) runs between Chicago and Seattle, the two largest electricity consuming areas in the BPA and MISO regions respectively, and runs through the heart of otherwise stranded, high quality wind resources in North Dakota and Montana. And pursuing an integrated approach could well allow the BPA to tap into the otherwise stranded resource in North Dakota and Montana at a lower incremental cost for the transmission capacity than the cost of a standalone transmission corridor. Similarly, an HVDC transmission corridor along the Union Pacific Southern Transcontinental corridor could connect the CAISO and ERCOT region.

 

Conclusion

As always, the end of the Sunday Train essay is not the final word, but the invitation to start the conversation. Remember that any aspect of sustainable transport policy is fair game for conversation ~ though having spent some hours writing about this topic, I may tend to force fit some other topic into this one, so please flag that you are raising a different topic when doing so.