We need more research on storage trading

This is one of two spin-off articles from my previous series on the rising importance of storage in electricity markets (available here, if you would like context and need some convincing).

Whether or not you are convinced storage will become the most important resource in every renewables-rich power grid, it is still important to understand how storage resources behave and how they alter wholesale market prices and outcomes (for good and ill) as we retire more thermal generation. While we already have examples of hydro and pumped hydro bidding as a practical demonstration, the introduction of a plethora of short-duration batteries and their auto-bidding algorithms might alter market dynamics and outcomes in unexpected ways.

The more solid steps in the supply stack are being removed

Wholesale electricity markets have always signalled the relative scarcity of electricity, to greater or lesser extent, according to their design. They largely do this via a supply stack resembling a staircase, each step representing the fuel cost of a thermal generating unit in a suitably competitive market. Thermal generation offers submitted at their marginal fuel cost are exogenous because those prices are determined outside the power system. This makes them more ‘solid’ than the offers of storage resources, which tend to make ‘shadow’ offers around the wholesale price being set by a thermal generator offer.

As the energy transition deepens, these ‘solid’ steps in the supply stack disappear as thermal units retire and the offer stack essentially becomes more like a slope – a sliding scale or barometer of market conditions indicating the likelihood the remaining capacity of available storage can deal with them. Think of the storage in the power system as one giant battery – prices rise as storage is depleted and rise faster as the likelihood of being short of generation increases. Similarly, prices fall as stores of energy increase and can fast approach zero or lower if stores of electricity potential become saturated.

Research is needed to understand how this changes market dynamics

What we want to understand is how the market is adapting as more and more storage resources enter and each of the solid steps in the supply stack are removed over the next decade or more. Note that renewable gas, hydrogen and derivative fuels become part of storage resources given the limits of their capacity.

Please check out my previous articles on storage for references to academic articles that consider the theoretically ideal and more realistic levels of alignment between private and societal objectives under imperfect competition, which has been added to by more recent work by Guillaume Roger from Monash University (Integrating Energy Storage into the NEM).

However, to understand better how market dynamics are changing, we also need more research examining what is actually going on in the bidding and price discovery process. A good example of this type of research is Abi Prakash‘s paper “The scheduling role of future pricing information in electricity markets with rising deployments of energy storage: An Australian National Electricity Market case study”.

…and to improve price discovery

We could also use more research like this to develop monitoring techniques that can check bidding and price discovery is working as we expect from a workably competitive market – something I imagine the market bodies are interested in. Further research and monitoring could help provide evidence and support for rule changes that improve the price discovery process and the level of competition in that process.

For example, there might be useful improvements we can make to the information market participants already have to make better decisions about the value of their remaining storage. This is important because, as the amount of storage increases, the value of electricity is increasingly a function of the marginal value of the electricity remaining in all the energy storage assets in the power system. On this front, I note that AEMO intends to publish the total state of charge of scheduled battery assets in each NEM region by 1 July 2027.

It might also be prudent to research the potential for humans and machines to find unique but counterproductive ways to attain their trading objectives. Collectively, auto-rebidding algorithms already have scale sufficient to affect outcomes in the NEM. There are plenty of examples where learning AI comes up with odd ways of solving problems (e.g. AI learns to play football, and breaks physics). Are we sure they cannot act in ways that threaten confidence in it?

Algorithmic share trading is regulated to address concerns about dark liquidity, high-frequency trading, and other practices that could distort price formation or create unfair advantages. Trading participants are required to report algorithmic trading strategies and monitoring systems to ensure fairness and prevent unfair advantage. They are also required to have robust risk management systems, including “kill switches” to shut down problematic algorithms, and to test the impact of their algorithms on the market. There is nothing like that for auto-rebidding software in the NEM.

Finally, I’ve already made my own short list of regulatory suggestions in a previous article.