Saturday, November 30, 2013

The EOS grid-storage system and nuclear power: a marriage made in heaven

EOS Energy Storage is peddling a megawatt-scale, fully containerized energy storage solution based on zinc-air (or zinc-oxygen?) cells. Self-contained in a standard 40-foot footprint, the cutaway shows blocks for batteries, inverters, and cylindrical objects which seem likely to be some sort of gas storage or perhaps filtering/processing system. The stated performance figures:

$1000/kW
$160/kWh
75% round-trip efficiency
Cycle life 10,000 cycles
30-year design calendar life

This appears designed to operate roughly 1 cycle a day for 3 decades.

If they actually deliver at those specs, it's worth thinking about what it could do. For instance, at $1000/kW output and 75% round-trip efficiency, $300 million invested plus 2400 MWH input over 6 hours (400 MW) yields 1800 MWH output (300 MW) over 6 hours.

Let's try this as a hypothetical example with something else that's already coming: the AP1000, with 8 currently being constructed worldwide. This will supply base-load power which can be cycled to follow load, but is most economical if it's run flat-out. The AP-1000 is rated at 1154 MW(e), and the estimated pricetag per plant of about $8 billion at Vogtle isn't out of line for first-of-a-kind efforts.... batteries not included.

Let's add them. $300 million for 300 MW of EOS units bumps the pricetag to $8.3 billion. Charging at full power drops the net output from 1154 MW(e) to 754 MW(e). Maximum discharge increases the net output to 1454 MW(e), nearly twice the minimum. (This is considerably greater than the 1.67:1 day/night swing detailed for the eastern provinces of Australia.)

At full cycling, the daily output is (1154*24-600)=27096 MWh, or 1129 MW(e) average. Other attributes:

Peaking: self-supplied (either centralized or distributed)
Reactive power: presumably available from the EOS inverter systems, distributed with the storage units.
Air emissions: zero.
Spinning reserve: as much as 700 MW (the difference between 400 MW maximum charging rate and 300 MW maximum discharging rate).

Amortizing $8.3 billion over 20 years at 7% interest costs $772 million/year; divided over 1129 average MW at 0.9 capacity factor, I get 8.7¢/kWh. Selling off-peak power at 5¢, mid-demand at 9¢ and peaking power at 15¢ I calculate $913 million annual revenue vs. $772 million annual amortization (salaries and fuel not included). Even at the extreme first-of-a-kind price of $8 billion for the nuclear unit, this is clearly affordable. After 20 years the bonds are paid off and the system becomes a cash cow for likely 4 more decades or longer.

The value added by the battery is the difference in purchase (or opportunity) cost of the off-peak power and the sales price of the peaking power. At the same 0.9 capacity factor I see $49.3 million annual gross revenue from the battery, paying off in just over 8 years. Plainly the battery is pulling its fiscal weight! But it will also cut the supply of off-peak power (shifted to charging), so off-peak prices may increase. This would further improve the economics of the system as a whole.

The impact on unreliables

Would the EOS battery make the dream of an all-renewable grid possible? That's very doubtful, given the need to tide the system over lulls adding up to days of average output. 48 hours of storage would itself cost $8000/kW, or around 16¢/kWh even if it was cycled continuously (50% capacity factor). That's over and above the cost of the power to charge it, which is hardly cheap at feed-in tariff rates. What would people do, looking at that pricetag to go "green"? They'd go the way of Germany and Poland, and burn coal. If stored energy comes at caviar prices, we should not be surprised if people decide to eat energy "junk food" instead.

The impact of a carbon tax

Suppose for a moment that the current system of production and investment tax credits is replaced by a simple, non-discriminatory figure of merit: a straight-up carbon tax. Let's set this carbon tax at $40/ton of CO2, which matches the 2.2¢/kWh PTC for a gas-fired generator emitting 550 gCO2/kWh. Coal plants will be assumed to emit 900 gCO2/kWh, with coal at 15 million BTU and $100/ton delivered (average bituminous and sub-bituminous). Also, with the North American shale-gas investment bubble about to collapse and multiple LNG export terminals ready to push prices up to world levels, wholesale NG delivered to major markets costs $15/mmBTU.

This was worth working through in detail, so I posted the spreadsheet in both text and downloadable file at ergosphere.wordpress.com. This spreadsheet assumes a grid capable of delivering 600 GW average, to allow expansion for electrification of transport etc. I used a 20-year amortization for all RE generation (wind farm lifespan appears to be shorter than that), 30 years for nuclear (licenses are now being extended to 60 years), 7% interest rate, and highly decentralized and interconnected networks for both wind and solar generation. Without storage the RE must be consumed at the time of generation, so transmission capacity must equal peak generating capacity. I assumed cost of $2 million per mile for a ±1.2 megavolt, 1000 A (2.4 GW) dual-circuit HVDC line with an average of 1800 miles length between generation and market. That's enough to get Dakota wind power to the coasts, and Arizona and New Mexico solar power to both Seattle and Georgia. I also rolled in a $40/tCO2 carbon tax for the fossil-backed options, with emissions of 550 g/kWh for gas and 900 g/kWh for coal. In the all-RE case, some 2.3 million miles of HVDC line are required. Some of these may be able to share rights-of-way; some may not. This many times the total mileage of the Interstate highway system. I assumed for the sake of simplicity that fossil-backed RE could use DSM to use peak generation productively and would require neither storage nor spillage.

The cost figures for the RE options are all dismal. Gas-backed is cheapest at $114/MWh (11.4¢/kWh), with coal not far behind. The gas option emits 122 gCO2/kWh, which is at least twice what we can tolerate in the long term. Getting this down using storage is staggeringly expensive. Using the EOS zinc-air system at $167/kWh, total cost soars by a factor of almost 10 and power rises to a prohibitive 90¢/kWh.

The nuclear option comes in best. Assuming $5000/kW average for a new-build fleet of nuclear reactors (roughly twice China's cost for a new AP1000), and 180 GW (1200 GWh) of EOS battery storage, total capital cost is about $3.3 trillion. No HVDC network is required. Amortization over 30 years at 7%/year is $270 million. Total amortization cost comes to 5.1¢/kWh. Carbon taxes are zero, so the only unknown is O&M at perhaps 2-3¢/kWh. CO2 emissions from operations are ZERO.

The nuclear system does not depend on natural energy flows, so it can be expanded when and where desired. For each new application electrified on this grid, all the carbon it formerly emitted is displaced. This appears to be a cost-effective way to de-carbonize entire national economies.

This would be anything but a small task. 632 GW of AP1000's is 575 units, not allowing for refueling and repair outages. Even so, building 30 a year the USA could finish the job in 20 years. The alternative is to build something like SMRs, where we'd be turning out several a week instead of one every couple of weeks. That looks doable too.

Conclusion

Trying to de-carbonize the US grid with enough excess to electrify transportation is a massive task. The cost of the all-renewable scenarios for doing it, with the requirements needed to provide a reliable supply to dark/calm parts of the country, is prohibitive. Nuclear energy and the energy stockpile of fissile metals eliminates both the long-distance interconnections and massive storage needed for reliance on fickle energy flows. If we want to go green, nuclear is the only real option we have.

Labels: alternate energy, analysis, batteries, CO2, energy substitution, nuclear power, petroleum dependence