by Rud Istvan
From the utility grid perspective, a fundamental problem with wind and solar is intermittency.
In the US, wind has a median capacity factor of 31%. In California’s Mohave Desert, solar PV has a capacity factor of 23%. To make up the electricity supply difference during the rest of the time, grids must either add otherwise unnecessary backup generation, or flex base load generation (dropping below optimum output so the grid can accept the intermittent renewable input). At a minimum, flexing results in costly capital inefficiency. Otherwise unnecessary backup generation is even more costly.
The higher the renewable penetration, the greater this intermittency burden becomes. For Texas’ ERCOT grid with 10.6% wind, the additional costs are ~$19/MWh for generation plus ~$6.50/MWh for transmission. It is now so expensive in Germany (26% renewable generation) that its largest utility, RWE, took a €3.3 billion impairment charge 1Q2014. The second largest, E.ON, took a €4.5 billion impairment charge 4Q2014, and announced it was spinning its conventional generating assets off into an unprofitable separate company. E.ON will also be shutting Irsching 4 and 5, large efficient CCGT units completed in 2010 and 2011! Irsching simply is not viable without being compensated for the forced Energiewende flexing it endures, while selling its electricity against the subsidized renewables with which it is also forced to compete.
So renewables advocates hope for major advances in grid storage to offset wind and solar intermittency. This guest post surveys what might be possible in the future given what is presently known. The focus is on utility scale, but takes an irresistible detour through TESLA’s newly hyped residential Powerwall. Sandia has a more detailed (albeit somewhat dated and hopefully slanted) utility storage analysis than this post, for CE denizens interested in digging deeper.
There are in principle only five ways that generated electricity can be subsequently ‘stored’: potential energy (e.g. pumped hydro), kinetic energy (e.g. flywheels), electrostatic energy (capacitors), electrochemical energy (batteries), and chemical energy (e.g. water hydrolysis). Anyone inventing another is in line for an automatic Nobel Prize (probably two, physics or chemistry plus peace).
Pumped hydro storage (PHS)
Potential energy in the form of pumped hydro storage (PHS, essentially reversible hydroelectricity) is >99% of existing grid storage worldwide.The figure is from EPRI 2011. EPRI has not updated their overall grid storage analysis, but did estimate that ~140,000MW of PHS was installed by YE2014.
All that a grid needs are upper and lower water ‘reservoirs’ in ‘hilly’ terrain, and reversible hydroturbogenerators. It is possible to excavate a lower reservoir deep underground, but at much higher cost. One such facility is proposed for Holland. O-PEC would be 1400MW x 6 hours for €1.8 billion, using a 1400m (!) hydrostatic head to minimize water and underground chamber volume.
Round trip efficiency is high at over 80%, and facility life is long at much over 40 years. The LCOE depends on facility size, cost, and hydrostatic head, ranging from as low as ~$85/MWh (EIA) to as high as ~$150 to $200/MWh (Sandia). PHS has so far been used mainly for peak load shifting. Off peak base load is used to pump water into the upper reservoir, which then generates back into the lower during peak load. This allows a larger grid proportion of low cost base load generating at optimal output 24/7 than would otherwise be possible. Where grid/terrain possible, PHS can also support renewable intermittency as already done somewhat in southern Germany, Austria and Switzerland (the Alps).
PHS always pays on a grid system basis if suitable affordable terrain is available. Many places favorable for wind (low relief Iowa, Denmark, northern Germany) or solar (low relief Mohave Desert) are distinctly NOT favorable—even though PHS does not have to be co-located, just very strongly grid intertied.
Developed world grids have already taken most of the advantage they can of PHS. California is an odd exception.
Kinetic Energy Storage
Kinetic energy storage is also extensively used on the grid, in the form of synchronous condensers for reactive power compensation (aka volt amp reactive, VAR). These are essentially unpowered generators spinning on grid. Some old decommissioned coal generating plants repurpose the old generators as ‘new’ synchronous condensers. Rotors may weigh hundreds of tons and spin at up to 3600 rpm, but still only store enough kinetic energy to provide transient voltage/frequency regulation (VAR). Downtown Tokyo alone uses six Toshiba purpose built 200MVARs. The human figure illustrates the enormous size/mass of grid scale kinetic storage machines. They are still grossly insufficient for bulk intermittent renewable storage.Beacon Power enhanced kinetic energy density by using grid-coupled carbon fiber flywheels spinning at 16000 rpm (reaching Mach 2, requiring they spin in a vacuum). Beacon’s first (subsidized) 20MW x 0.25hour facility comprised 200 flywheels and cost $4800/MWh. Its purpose was frequency regulation (VAR), not bulk energy storage. This facility is the flywheel capacity in this post’s initial EPRI energy storage figure.
That is far too little energy for bulk utility wind or solar intermittency. To back up a single 2MW wind turbine at 30% capacity factor would require ([1-0.3]*24 hours / 20MW/ 2MW * 0.25hour) 6.7 of the pictured facilities.
Electrostatic storage in capacitors is ubiquitous. There are large capacitor banks for reactive power compensation (VAR) on all utility grids.
Super capacitors have the highest energy density of any capacitor type, an order of magnitude more than the next best kind. Supercaps plus power electronics have created a rapidly growing new utility device class in the past decade, static compensators (statcoms). These substitute for smaller synchronous condensers; ABB’s statcoms come in sizes up to 30MVAR. Their advantage is no moving parts/maintenance. Like synchronous condensors and capacitors, statcoms store far too little energy for bulk wind and solar needs.
Electrochemical batteries presently have limited use on the grid. Rechargeable battery electricity is stored in some reversible electrochemical reaction. Familiar lead acid (PbA) electrochemistry is sponge lead/lead dioxide electrodes creating/removing lead sulfate, with sulfuric acid electrolyte conducting the needed sulfate ions. Which is also why deeply cycled PbA batteries have inherently short cycle life unsuited to utility storage. Cycling grows ever-larger and increasingly insoluble lead sulphate grains (sulfation), while the growing/shrinking lead sulfate in the electrodes eventually causes them to disintegrate from mechanical stress. Xtreme Power designed an industrial/utility PbA capable of 650 cycles to 80% discharge, which would last less than two years supporting solar. Xtreme delivered (to utilities) about 35MW at about $1000/MWh before going bankrupt. Xtreme is the PbA in the initial figure.
There are many reversible electrochemistries. Commercial ones include PbA, Nickel Metal Hydride (NMH, in most hybrid autos), lithium ion (LiIon, ubiquitous in portable electronics and electric vehicles), and sodium sulfur (NaS). There are several ‘experimental’ chemistries with one or more as yet unresolved issues. These include lower cost lead carbon (PbC, cycle life), lower cost and higher energy zinc air (cycle life, safety), lower cost and higher energy sodium ion (cycle life), higher energy lithium air (cycle life, safety), higher energy solid state LiIon (SSLiIon, cost, cycle life), and higher energy lithium sulfur (LiS, cycle life). Furthest along seem to be PbC, SSLiIon, and LiS (the links are illustrative, not exhaustive of all the entities working on these electrochemistries).
There is a lot of uninformed MSM reporting on battery progress, often based on hyped lab PR (most recently Harvard’s rhubarb battery, below). Electrochemistry has been known since Alessandro Volta’s 1799 stack of zinc, brine soaked paper separator, and copper twitched a frog’s leg. Sticking zinc and copper into a lemon still works—but not for any practical application. Many, many $billions have been spent on battery R&D over the past century. Progress remains a very slow slog. It is beyond unlikely that any fundamental electrochemical miracles remain unrevealed.
There are two basic battery design concepts. The familiar one (like PbA, NMH, and LiIon) stores electricity in the electrodes. This is not a problem for portable electronic tiny energy storage needs. It is a big problem for utility bulk storage requiring a lot of expensive electrode. A123 Systems delivered one 20MWh LiIon system (10 of the imaged containers) that cost (based on its federal loan guarantee) $17.1 million. This facility is the LiIon capacity in the initial figure.
Backing up a single ~$3.5 million 2MW wind turbine at 30% capacity factor would require 8.4 of these containers at a cost of ~$14.4 million. They would be purchased from NEC Energy Solutions; A123 went bankrupt. Its assets were sold to the Chinese at a $119 million loss to US taxpayers who gave A123 grants and loan guarantees. The Chinese sold the utility portion to NEC. NEC has a new order for 3 20MW installations for the PJM grid. Like Beacon’s flywheels, these are for frequency regulation, not bulk renewable energy storage. In that application (which does not deeply discharge the batteries), NEC says they last 20 years—still insufficient for wind or solar without replacement.
If energy density can be increased, then the amount of electrode per unit electricity can be decreased proportionately. LiIon’s theoretical limit is 1015 Wh/liter. The Panasonic cells Tesla uses (below) are about 620 Wh/L (cells, not the Powerwall battery at ~350 Wh/L with liquid cooling). Panasonic produces more advanced (and more expensive, shorter lived) smartphone cells that approach 800 Wh/L. There will undoubtedly be some further improvement in LiIon energy density, but nothing like what has already been achieved.
The main ‘electrode concept’ utility battery is NaS. Several hundred MW are already on grid, as the first figure shows. These operate at 350C (and must be kept at that temperature continuously), have a 15-year life (significantly longer than deeply discharged LiIon at ~10 years per Tesla), and have a round trip efficiency of 75%. They are used in special grid distribution situations, for example to support remote peak loads where a small peaker or a beefed up transmission line would be even more expensive than NaS. California’s PGE just installed a 4MW x 6hour NaS facility from Japan’s NGK. It cost $18 million. At 30% capacity factor, about (16.8 hr *2MW/[4Mw x 6hr = 24MWh NaS] /0.75 efficiency) 1.9 of these facilities would be needed to back up a single 2MW wind turbine. NaS would cost $34 million to back up one $3.5 million turbine. And it would have to be replaced after 15 years to support the turbine’s ~25 year life. NaS is not commercially feasible for renewable bulk electricity storage.
The other class of battery design stores electricity in the electrolyte, and only uses smallish expensive electrodes to put it in and take it out as electrolyte is pumped through the electrodes. In such ‘redox flow’ batteries, the electrolyte can be stored in arbitrarily large tanks. This theoretically solves the grid scale electrode cost problem. And also part of the battery cycle life problem, since the electrolyte and/or electrodes can be separately replaced.
There are several redox flow chemistries in development. Some are expensive and corrosive, like the vanadium redox battery (VRB). The most recently hyped experimental system uses inexpensive organic quinones similar to those found in rhubarb. No word yet from Harvard on round trip efficiency, or how long their ‘breakthrough’ rhubarb flow battery might last.
One seemingly promising type (since a commercial unit exists) uses relatively (compared to vanadium) inexpensive iron/chromium, championed by 2008 California startup EnerVault. Their first ‘commercial’ flow battery (250kw x 4hour = 1MWh) was installed in 2014 to support a solar powered 250kw irrigation pump in Turlock, California. It has a round trip efficiency of 60%. To back up a single 2MW wind turbine at 30% capacity factor would require (16.8 hours *2MW/ 1MWh/0.6 efficiency) 56 of the pictured EnerVault facilities. Or requires tanks, electrodes, and pumps that are 56x bigger than pictured.
The facility cost $9.5 million, $4.7 million from a US grant. EnerVault told EPRI and DoE (as part of the grant process) that it expected to be ~$350/kWh in volume production. Backing up a single $3.5million 2MW turbine might cost ‘only’ $12 million in the future. The useful lifetime is TBD; EnerVault says >20 years. But EnerVault also says the pumps last “thousands of hours” before needing replacement. That could mean yearly—and probably does. On April 14, 2015 EnerVault announced it was ‘restructuring’ (laying off most employees) and ‘seeking new owners’. Existing investors including Japan’s Mitsui, French oil company TOTAL, and 3M declined to put in more money. EnerVault has failed.
The foregoing examples illustrate the immensity of the utility bulk storage challenge. No foreseeable battery solution overcomes this enormous challenge.
Distributed grid storage
There has been much renewables discussion of ‘distributed’ grid storage. Put many smaller batteries at residential or commercial locations, in the hope that manufacturing volumes would provide cost economies of scale. Thus the MSM excitement Elon Musk created with his Tesla Gigafactory and the Powerwall. The 7kWh daily cycle unit (complementing rooftop PV) has a guaranteed 10 year life at 92% round trip efficiency for $3000, excluding installation and inverter.
Whether Powerwall makes any sense is a less exciting question, which Musk’s fawning MSM did not ask. Palo Alto’s approximate LCOE for rooftop PV is ~$0.155/kWh for a 5kWDC system before subsidies, according to Palo Alto itself. To charge a single Powerwall while still using the original PV as before, about (7kWh/5.4 ‘sun hours’ per NREL /0.92 efficiency) 1.4 kW of additional PV would have to be installed. Using Palo Alto’s ‘official’ estimate, that is an additional PV cost of about $8400 (including the inverter Tesla does not supply). Total cost $11,400 for one Powerwall is no problem—if you can afford to live in Palo Alto and install PV there in the first place. The city says its average home consumes about 1000 kWh/month or (1000/~30.5) 32.8kWh/day. A 6.4kW PV plus one Powerwall will not take an average Palo Alto home off grid—it is (32.8-12) 20.8kWh short. It would take 4 Powerwalls (plus their additional PV) to go off grid. Not enough dollars or roof to make that work.
Being a little bit Palo Alto/Tesla green comes at a large cost. The 10-year, 0.065 discount rate LCOE of single Powerwall is $0.118/kWh. To that must be added the LCOE of the extra charging PV, adjusted for Powerwall efficiency. According to Palo Alto, that is (~0.155/0.92) $0.168/kWh, for total Powerwall LCOE of $0.286/kWh. The residential cost of electricity in California (March 2015) averaged ~$0.17/kWh. Powerwall is a bad deal, costing almost twice what California’s residential grid electricity does. (Tesla cars are a similarly bad deal.)
Hydrogen can certainly be hydrolyzed from water. And the necessary electricity can certainly come from intermittent renewables. The most efficient way to convert hydrogen back to electricity at grid scale would be a PEM fuel cell or an SOFC. The math can be done using Ballard’s 1MW PEM, since a few have actually been sold as demos. Ignore the technical difficulties of bulk hydrogen storage, which the following methane alternative ‘solves’.
The theoretical efficiency of hydrolysis is ~88%. About 4% of commercial hydrogen is made this way today, with real efficiencies of ~75%. EERE says PEM fuel cells can be 60% efficient. But that is also theoretical. Ballard’s real 1 MW ClearGen® is 40±2% efficient, with a lifetime of ~15 years (similar to NaS). The round trip efficiency of a hydrogen electricity storage system would be about (0.75 * 0.4) 30%. For a utility, that is awful.
The electricity to be stored comes mainly from otherwise flexed base load generation, with chemical storage buffering renewable intermittency no different than PHS buffers peaks. The energy cost alone would be about ($57/MWh baseload / 0.3 efficiency) $190/MWh. Ballard’s ClearGen® costs about $10 million/MW (including inverter, transformer, and installation).That calculates a capital LCOE of about $114/MWh. Adding hydrolysis and H2 storage, the system LCOE is >>$304/MWh. It is simply not commercially viable–by nearly an order of magnitude. Before solving the hydrogen storage problem.
Or, hydrogen from electrolysis could be reacted with CO2 over nickel catalysts to produce methane. Methanation is significantly exothermic, although up to half of the resulting ~20% ‘waste’ heat could be reused (e.g. heating input feedstock, since the catalysis works between 200C and 550C). A number of lab scale reactors plus at least one pilot facility have been built, with methane yields from ~70% for one pass to ~95% for three. Solves the hydrogen storage problem. The resulting synthetic methane can be stored and used like natural gas from any other source (e.g. in flexible CCGT with 58-61% efficiency). Input CO2 could theoretically be obtained from fossil fuel carbon capture (without sequestration), a process with 20-30% parasitic energy loads.
Methane round trip energy efficiency would be about 0.75 (hydrolysis) * 0.95 (catalysis yield) * 0.9 (net [half] methanation exothermic loss) * 0.8 (minimum CC parasitic load) * 0.6 (flexed CCGT) or ~31%— no better than hydrogen alone, after all the chemical complications. Methane storage avoids the technical hydrogen storage challenge, but at the expense of much additional chemical plant capital, operations, and maintenance cost. It worsens the chemical storage economics.
Most renewables advocates don’t appreciate the scope and scale of electricity grids, the difficulties intermittency creates, and the technical/ commercial inadequacies of electricity storage technologies other than PHS.
Utilities already utilize four out of five forms of energy storage wherever they make sense. Potential energy is ubiquitous pumped storage. Kinetic energy is ubiquitous synchronous condensers. Electrostatic energy is ubiquitous capacitors and statcoms. Conventional electrochemical batteries are not practical except in special situations, and probably never will be. Flow batteries may improve on conventional batteries somewhat, but are still far from feasible for large-scale bulk wind and solar storage needs. Chemical storage is even worse than batteries because of its inherently greater thermodynamic inefficiency.
It is very unlikely that any grid storage solution (other than PHS where feasible) could ever practically cover the intermittency of high penetration utility scale wind and solar. Utility voices (like RWE and E.ON) charged with making electricity grids work seamlessly and reliably despite ever increasing renewable intermittency burdens are only starting to be heard. Those voices are very negative. It may not be until some grid goes dark because of intermittency (as increasingly uneconomic flexed conventional generation is shut in Germany and UK) that the general public will understand. Germany, UK, and California seem determined to run this unfortunate experiment for the rest of us. One or more appear likely to succeed soon in experimentally proving the grid instability ‘blackout’ hypothesis. The question is mainly when, not if.
Acknowledgements. APlanningEngineer provided many helpful comments on earlier drafts.
JC note: As with all guest posts, keep your comments relevant and civil.