The Untapped Potential of Grid Storage
Electricity is unlike any other commodity. It is invisible. It is consumed almost immediately after it is generated. And it's almost entirely fungible.
To make it fit market mechanisms, we develop convenient abstractions. We say we're buying a kilowatt-hour from a generator, even though without our load, there would be no kilowatt hour to buy. We say we're "feeding-in" our solar to the grid, when we're actually feeding a tiny part of all the loads connected to the grid. We say we're purchasing "green" electricity, when as far as electricity consumed is concerned, there's no distinction by colour or creed.
These abstractions promote a mental model of electrons being pushed in at one end and pulled out at the other end. In fact, a more accurate model is a stiff iron bar, with the generators banging one end at a constant frequency, and everyone else connected to the bar riding the vibrations. Nothing material flows from generator to consumer, only energy. And it's impossible to tell whose banging you're riding.
These abstractions work pretty well, and energy markets thrive. But when the world contemplates a shift from 100 years of electricity coming from material fuels, it's worth re-assessing them. For a hundred years we relied on the biggest solar battery, the earth, as our fuel stockpile. Now, in many parts of the world, we're cutting out the middle man and soaking up what the sun lays down on us each day, in insolation, wind and waves.
There's a dramatic distinction in this approach - this type of fuel comes and goes whether we use it or not. For a functioning solar or wind generator, the cost to capture a kilowatt hour of electricity is the same as the cost to capture nothing. Instead of fuel making up the majority of the cost to generate, there are only less variable costs like capital and maintenance.
Near zero fuel costs, particularly for renewable energy rich nations like Australia, opens up enormous potential. The greatest mistake we could make is to treat it like traditional generation, and miss the upside.
Regardless of their cost structure, generators need to get paid. So the market operates on a fungible unit - the dollar per megawatt-hour ($/MWh). The more energy you supply, the more you get paid, up to the market demand. The business proposition looks very different for a coal plant, with their fuel supply contracts and their difficulty ramping down, compared to a solar plant, with no fuel costs but an inability to supply at night. But for the most part the $/MWh mechanism is both fair (investors are rewarded for building generators) and competitive (the market will favour cheaper forms of generation).
The mechanism even extends to grid storage, which in itself does not generate electricity (in fact, there is always some lost via storage). That's because when generators with low fuel costs have supply that exceeds demand (or system constraints), they can afford to offer it at very low prices. Similarly, when they are unable to meet demand, the remaining suppliers can command very high prices. The case for storage to play arbitrage - buying low and selling high - is obvious.
Take a step back however, and you can see a flaw in the system. More and more generators have zero marginal cost because they get their fuel for free. Their generation is no longer tempered by fuel costs, but by the limits of serviceable demand (this is, the demand that can be reached via the system). But since energy from generators of this type is correlated in time (driven by the same weather or diurnal patterns), their constraints compound. That means incentives to build new generation is limited by the constraints at the very times they're best able to supply. Despite there being unmet demand at certain times and certain locations, investors are not rewarded for building more. Meanwhile, as arbiters, storage can only react to emerging constraints. Only when those constraints become apparent (or can be safely foreseen), does a market opportunity arise. There is no arbitrage opportunity in creating constraint headroom, or being too early with relief. And without headroom, new generation immediately meets constraints, resulting in an uphill battle for return on investment.
We're starting to see this doldrum play out in the NEM. When variable renewable energy (VRE) generation arrived at scale, the arbitrage opportunities became very apparent. Wholesale electricity price began to fall below $-1000/MWh and spike above $10,000/MWh more regularly. Grid scale batteries swooped in and started to make bank on the spread. But instead of swinging the pendulum back the other way, the volatility has sort of just stuck around.
The problem is that battery operators too, must get paid. And if your revenue comes from arbitrage, your goal is to ride it, not eliminate it. A battery storage system only lasts so long, and since it doesn't cost you any extra to run the plant at 100% compared to sitting idle, your best bet is to cycle as much as possible. Predict the bottom and top of the market price every day (or twice a day if you can!) and take what you can. Since most grid batteries are running a similar playbook, they tend to correlate in time too, charging during predictable falls in price and foregoing unexpected price falls because they're full. Then discharging during predictable price rises, foregoing unexpected or future price spikes because they're empty.
The net effect is that battery installations rise to the point where they start eating each other's lunch. Unfortunately, that's before headroom is generated that would incentivise more generation. As wholesale prices come down, there's even less incentive to build generation. Average revenue at two of Australia's biggest solar farms fell to $39/MWh last year. Earnings fell even as output rose. These plants have energy to give, but find themselves in a market that rewards them less and less.
Left unchecked, generators and storage provides will find themselves nibbling from a smaller and smaller cake.
There are efforts underway to find better models. Some battery operators eschew the "merchant" approach. They prefer to demonstrate their long term potential as key components in energy supply, entering into Power Purchase Agreements instead. And on the market operator side, AEMO have introduced a "System Integrity Protection Scheme" (SIPS) reminiscent of the "Reserve Capacity" incentives in the West Australian grid. It rewards storage providers for simply having some energy "ready to go".
- https://theenergy.co/article/making-the-most-of-big-batteries
- https://www.aemo.com.au/newsroom/media-release/aemo-completes-system-integrity-protection-scheme-procurement-process
There is also a concerted effort to champion and undertake major transmission projects. Transmission has an essential role to play in alleviating (or more favourably, capitalising on), the existence of excess generation afar from the demand. But across the world, transmission is proving extremely difficult to build. Who would have thought ferrying electricity back and forth by rail would work out to be an economical alternative?
But if we allow ourselves to take a systematic view of the energy network, then it becomes clear storage has a bigger role to play than just arbitrage. The existence of storage, distributed around the grid, would provide reliable demand for new generation. New generation would be free to generate where and when it made sense, and get paid for it (minus a storage fee perhaps). Free of the need to maximise wholesale price spread, storage could redistribute that energy to minimise impact on congestion. It could for example, transfer the stored energy to a storage system closer to load centres, but at a time of low transmission utilisation.
If we continue to apply the systematic view, storage need not aim to discharge completely during the daily price surge. Instead, for the good of the grid, it could ensure there is always some in reserve, ready to discharge "green energy" to meet unforeseen supply constraints.
Looked at this way, storage fulfils the potential of renewable energy generation. Instead of treating VRE as always a bit too much or far too little, compensating with price volatility, we treat VRE as the predominant source of energy. Storage just works to spread it out.
Rather than forcing project developers to fight over smaller pieces of the pie, storage grows the pie. With the bridle of fuel costs alleviated, aspiration is constrained by the appetite of capital rather than the burden of preserving finite stockpiles. Energy is perhaps one of the best examples of Jevon's Paradox - making it more efficient tends to be outweighed by the additional ways people find to use it.
If we do choose the aspirational view, what is the correct way to price storage? I'm not sure, but given it's irrelevance to the costs of running a storage system, $/MWh sure sounds wrong. It's sort of like valuing rest stops on the highway by how quickly vehicles come and go, rather than by how they spread out laid up trucks, or make long trips safer and more enjoyable. By the throughput metric, a rest stop design would maximise utilisation, perversely undermining the very thing that makes them valuable - availability and dwell time. Rest stops don't provide traffic, they provide respite, which makes the roads more effective and more popular.
Oddly, the $/MWh carrot incentivises storage in the same way it does generation - to produce energy transactions as fast as possible. Indeed, generation is rated in terms of the rate it can produce energy: megawatts (MW). So storage plays an interesting game - it exists to allow energy at rest (MWh), but is paid for energy in motion (MWh per hour, ie. MW).
For grid scale storage, one of the unit costs that does materially apply, is capital utilisation. Grid scale batteries have plummeted in price recently, but they remain bounded by number of cycles. So the return over the life of the plant is relevant. At an all-in installation cost of, say, $200/kWh, a battery playing the arbitrage game might hope for an average $100/MWh spread. So to break even it would need to cycle 2000 times, regardless of size. But a dollar now is worth less than a dollar in the future. Plus, the current iteration of battery technology comes with a use-it-or-lose-it guarantee. Performance degrades over time, so operators are doubly incentivised to make the return as soon as possible. Practically speaking, a modern grid scale battery can cycle between one and two times per day. At that rate you can get 2000 cycles done in under 5 years, and start to make good on your investment until your equipment starts to degrade anyway at four or five thousand cycles.
If storage is to fulfil its potential, it needs to provide value even in the absence of price spread. To do so would require breaking the rush to return on investment. That's what makes Long Duration Energy Storage (LDES) so interesting. It's not so much that batteries cannot store their energy for a long time, it's just that their economics discourage storage at rest. So LDES might be better thought of as Long Living Energy Storage. In other words, enabling infrastructure that exists to make the system more effective. It's goal is to complement Short Living Energy Storage, which seeks out the most predictable price spread every day, by instead providing a permanent assurance that generators can generate at the most cost effective times.
Get storage right, and we have the potential to bridge space and time, turning low fuel cost energy generation from occasional grid participant, to driver of demand we haven't even invented yet.