‘At least six months ahead of go live’: OptiGrid on the ideal onboarding runway for battery storage systems

“All batteries have warranties and long-term service agreements that they need to take into account when you’re optimising their trading,” he says.

“Sometimes those constraints can actually change the trading behaviour in a way that would result in weird outcomes.”

Karimi points to South Australia’s price cap event on 21 June 2026 as an illustration of what can go wrong when those constraints aren’t properly integrated into a battery’s dispatch strategy.

Prices in the state’s SA1 region hit the National Electricity Market’s (NEM) AU$20,300/MWh (US$14,654) cap twice in a single evening, with the region’s fleet of grid-scale batteries capturing a combined AU$324,000 in revenue, though performance across individual assets varied sharply.

Karimi previously told ESN Premium that the divergent outcomes among the state’s 15 grid-scale battery storage systems came down largely to state of charge management and bidding strategy rather than raw capacity, with some assets capturing meaningful revenue while others were caught charging into the cap itself.

Autonomous optimisation over manual override

Karimi is direct about how he believes constraints should be handled operationally, arguing against a model in which human traders periodically disable automated optimisation to intervene manually.

“We don’t think the optimal way to operate the battery is to turn off the optimiser and then do manual bidding and then turn it back on,” he says.

Instead, Karimi describes an approach in which human input shapes the parameters within which an optimiser operates, rather than replacing it.

“What’s optimal is that you allow the human trader operator to input their preferences, their constraints, their objectives, and then the optimiser should automatically take those into account and then optimise the revenue within those bounds,” he says.

He notes that OptiGrid’s platform, OptiBidder, does allow human traders to adjust the optimiser’s behaviour, but the design philosophy keeps the algorithm running continuously rather than ceding control entirely during periods of manual adjustment.

That framing extends the argument Karimi has previously made about the gap between a battery storage system’s theoretical earning potential and what it actually captures through trading decisions, a gap that metrics such as normalised revenue and the percentage of perfect foresight are increasingly used to measure across the industry.

Constraint-aware, always-on optimisation, in his account, is one of the main levers for closing that gap.

A six-month runway before go-live

Asked how far in advance of a project’s energisation OptiGrid typically begins meaningful optimisation planning, Karimi says the answer depends on the asset owner’s own requirements but sets out a rough internal benchmark.

“From our perspective, we need at least six months ahead of the go-live,” he says, describing that window as covering integration work with a project’s SCADA control system in parallel with setting up the trading strategy.

During that period, the optimiser runs in a digital environment, allowing human traders to observe its performance and test their own strategies before the asset actually starts trading.

Karimi says OptiGrid has worked to shorter timelines than that on occasion, but that more lead time generally produces a better outcome.

“The more time we have, the better, in the sense that we can make sure that we’ve ticked all the boxes,” he says, “and everything is done in an optimal way and with enough time to basically make sure all the integration, trading strategies, and onboarding is done properly.”

When asked how OptiGrid mitigates the risk of correlated dispatch as its network of optimised assets expands, it was noted that battery storage systems optimised by the same platform could potentially exhibit similar bidding behaviours. This could lead them to compete for the same price-setting opportunities within a dispatch interval.

Karimi argues that the risk is manageable because optimisation outcomes depend on each asset owner’s specific constraints and objectives, not solely on shared market and price forecasts.

“Even with the same price forecast, even with the same market forecast, the behaviour will not necessarily be the same because they’re following different objectives,” he says.

He notes that OptiGrid has observed this directly within its own portfolio: “Even we have seen, even on our optimiser within the same region, the optimal bids are different depending on how the optimiser is set up,” a variation he attributes to the onboarding process, during which each asset’s constraints are configured individually.

Karimi also discloses a second product in development, OptiTrader, a portfolio-level optimisation and risk management layer designed to sit above OptiBidder for asset owners managing multiple battery storage assets in a single market.

He says the tool is intended to prevent one asset’s dispatch from cannibalising the revenue of another held by the same owner.

“That’s for companies that have portfolios of assets, ensuring that each asset will not cannibalise the revenue from the other,” he says, distinguishing that risk from cross-owner correlation.

Because different owners typically bring different constraints, offtake obligations and trading strategies, Karimi says their assets are less likely to behave identically even when optimised on the same underlying platform, whereas assets under common ownership face a comparatively higher risk of similar, and therefore self-cannibalising, dispatch behaviour, the specific problem OptiTrader is designed to address.

Never miss an Australian energy storage story. Sign up for our Australia newsletter and get the latest project announcements, policy updates and market analysis delivered directly to you.