Percentage of Perfect

2025Q3

The battery revenue stack consists of energy arbitrage and frequency control ancillary services (FCAS) revenue. Unlike wind and solar projects, which largely generate depending on whether the sun is shining or the wind is blowing, batteries are fully dispatchable. Batteries actively decide when to charge and discharge to capture arbitrage opportunities and participate in the eight FCAS markets. Given the dynamic nature of the electricity market – with supply and demand constantly fluctuating with weather, plant availability and transmission constraints – the optimal strategy for battery dispatch can change from one five-minute trading interval to the next.

This introduces significant operational complexity. Batteries must continuously reforecast the price outlook to identify optimal market participation. In practice, these decisions are typically managed by autonomous bidding optimiser software, which dynamically evaluates market conditions to maximise revenues across arbitrage and FCAS services and automatically generates market bids. For developers and asset owners, optimiser performance plays a key role in realising target returns from their batteries.

While everyone in the battery industry agrees that benchmarking battery performance is essential for evaluating software providers, there is still no single standard for assessing optimiser performance. Benchmarking battery performance is complex due to the range of industry methodologies. Simple revenue comparisons alone cannot capture project-specific factors such as battery duration (one-hour vs two-hour), site constraints, plant availability, contractual terms, or off-market incentives.

One of the methodologies that is commonly used to benchmark the optimiser performance is via a Percentage of Perfect (PoP) method. This approach involves back testing the battery’s performance within historical market conditions on a perfect foresight basis. In other words, if we knew the electricity and FCAS prices beforehand, how would we trade the battery in terms of charging, discharging and participating in FCAS markets. This gives us the maximum revenue the battery could have earnt over a given period, and then comparing it to the actual revenue earnt during the period. Formulaically, percentage of perfect can be written as:

Percentage of Perfect =Actual revenue earnt / Maximum revenue opportunity on a perfect foresight basis.

While, PoP provides a method of benchmarking optimiser performance, it comes with a few limitations. This approach assumes that the battery’s activities have no impact on market prices – it is a price taker and can dispatch all of its capacity at historically observed prices. This assumption falls short of reality as bidding behaviour will influence prices.  This is particularly the case for larger batteries or even for smaller batteries where they are collectively controlled by a single optimiser that controls a large amount of capacity.   This is important at extreme price events – e.g. when electricity prices are close or at the market price ceiling of $20,300/MWh.  At these times, by definition, there is very limited supply, and the actions of a single party can often have a material impact on prices.

For example, consider a scenario where peak demand and prices were expected to occur at 7pm, followed by a sharp fall in prices at 8pm when night-time wind generation is expected to arrive. Considering the forecast, all batteries in the market dispatch their available capacity at 7pm to capture the peak prices. However, let’s assume that the wind was slow to arrive and therefore resulted in a surprise shortfall of supply at 8pm. At that point, all batteries have discharged and have limited opportunity to respond to the shortfall and prices hit the market cap of $20,300/MWh. Under the PoP methodology, batteries should have foreseen the supply shortage at 8pm and withheld capacity at 7pm in order to capture the price cap at 8 pm. Yet in practice, if a significant number of batteries had done so, the shortage at 8pm would not have occurred. There would have been supply available to fill the gap left by wind generation and prices would not have reached the cap.

Another limitation of the PoP benchmark is that it can only be applied to merchant batteries. Batteries that have some form of contracting for providing network support or a part-tolling arrangement are constrained by obligations under the arrangement. Such batteries will not follow a 100% revenue maximising strategy. For example, the Victorian Big Battery and the Waratah Super Battery have contractual obligations with AEMO under the System Integrity Protection System (SIPS) protocol to ensure minimum availability of the battery at certain times of the year. Given the contractual requirements, it is hard to benchmark the battery on a pure PoP basis.

Similarly, for a hybrid solar and battery project, the calculation of PoP is even more complicated and may not provide a basis for benchmarking across different battery projects. A hybrid solar farm charges the battery from the solar panels, not from the grid. The decision tree for a hybrid solar farm has a few more branches as the project actively decides between the following options to maximise revenue:

• Charge the battery from the solar farm

• Discharge excess solar generation to the grid

• Discharge the battery to the grid

• Curtail (turn off) the solar farm and not dispatch any generation to the grid (eg if market prices are negative and the battery is full)

  
  

While theoretically, one can predict what the maximum revenue is, back-casting a hybrid is quite complex as each project comes with its own nuances, and you need to know what potential solar generation was (e.g. generation before economic curtailment for negative prices). Curtailment due to thermal constraints and negative pricing are different across projects which make benchmarking solely on the basis of percentage of perfect unreliable.

Despite its limitations, participants continue to use PoP for benchmarking battery performance. Typically, proponents of battery projects have assumed that the PoP for a project will be close to 80-90%. This plays a key role in their expectations of future revenue outcomes as revenue forecasts are calculated on the following basis:

Forecast Revenue = Market revenue x Percentage of Perfect

While there are no publicly available ratings for optimisers operating within the NEM, analysis of a sample of optimisers has revealed that actual observed PoP for optimisers have been more in the 60-80% range. There is significant variability between the PoP performance for individual optimisers. Some optimisers have been performing far better than others.

So, what does this mean for the future?

If an optimiser is chronically underperforming PoP expectations relative to its peers, battery project proponents are highly likely to switch their optimisers to better performing optimisers. We wouldn’t be surprised that the dynamics of funds management industry start to emerge in battery optimiser markets - where there is constant switching from underperforming managers (optimisers) to outperforming managers (optimisers).

The job of an optimiser is about to get harder as batteries become a dominant pillar of the generation fleet in the NEM. Currently, on average batteries make less than 2-3% of the total evening peak supply and primarily compete with coal and gas power plants to be dispatched. It is easier to optimise when the competition has a reasonably predictable cost of supply and, hence, bidding behaviour (fuel costs determine how coal and gas bid). However, this dynamic will change in a battery dominated world, where the competition shifts from a human trader responsible for dispatching coal/gas power plant to another algorithm optimising a battery. This is where we believe electricity markets cross over to high-frequency trading. Optimisers will need to consider bids from other batteries and predict bidding behaviour, similar to order flow predictions. The real alpha for an optimiser might not just be optimising for peak prices but predicting optimisation errors by other optimisers and front running them (algorithms competing against algorithms). Will this have a further impact on the PoP? Only time will tell.

Lastly, if the current levels of PoP in the 60-70% range are consistently observed in the future, it would be very hard to justify the 80-90% POP assumption that underpins many long-term financial model forecasts. Revising the PoP down, ceteris paribus, would be a material write down of revenues compared to expectations which will disappoint long-term infrastructure investors. Active management and proper benchmarking of the optimisers will be key to realising full value of battery projects.