News
Dispatch Is a Design Choice

Battery arbitrage is often presented as a tariff problem. In practice, it is a control problem.

Most deployed systems still rely on fixed percentile thresholds over rolling historical windows. If the price is in the lower quartile, charge. If it is in the upper quartile, discharge.

The logic is simple. It is also time-invariant.

A low price at 03:00 is routine. A low price at 18:00 is information. Conventional dispatch does not distinguish between the two.

Electricity markets are structured systems. They exhibit daily rhythm, seasonal distortion, regime shifts and, at times, persistent deviations from their own norm.

We examined whether incorporating those structural properties directly into dispatch logic—without forecasting future prices—produces measurable improvement.

The result is a adaptive, zero-lookahead framework that:

  • Learns diurnal structure from observed prices.
  • Separates structural pattern from residual deviation.
  • Estimates persistence using the Hurst exponent.
  • Modulates thresholds via fractional Brownian motion theory.
  • Introduces no tuned constants.
  • Runs on modest compute.

It does not require external forecasts, but derives its predictions from the structure already present in the price signal.

Empirical Results

The method was backtested across seven full calendar years (2019–2025) of DE-LU Day-ahead prices using a reference 13.5 kWh battery model.

Relative to fixed P25/P75 dispatch:

  • It outperforms in every year tested.
  • Uplift ranges from 2% in quieter regimes to nearly 30% during high volatility periods.
  • Divergence is strongest in persistent, stressed markets.
  • Absolute benefit scales linearly with usable capacity

The failure modes are visible and explainable. Summer solar distortion weakens the diurnal template; the effect is measurable and correctable. Control quality matters most when volatility expands. That is also when arbitrage value expands.

Image

Average DE-LU day-ahead price by hour of day, October 2018 – February 2026 (74,170 hourly observations). The bimodal shape reflects the interaction of demand patterns with solar generation: an overnight low, a morning demand peak, a midday solar trough, and an evening consumption peak.

When Simple Stops Being Sufficient

Fixed percentile dispatch is often adequate in low-volatility conditions. In those regimes, errors are small and rarely noticed. Spreads are narrow. The cost of imprecision is limited.

But electricity markets are not stationary. During periods such as 2021–2022, volatility and persistence increase materially. Under those conditions, structurally blind control produces widening performance gaps. As storage penetration increases, simple strategies converge. The obvious trades become crowded. Marginal performance shifts from asset capacity to control architecture.

The question is not whether percentile dispatch functions. It is whether it remains sufficient.

Image

Daily arbitrage earnings by strategy. Grey bars: blind deterministic. Blue bars: MF Predictive. MF captures more value on high-spread days and limits losses on low-spread days.

Where It Matters in Practice

In live markets, dispatch decisions determine whether stored energy is released into genuine scarcity or into routine evening uplift. They determine whether cycles are deployed against structural spread or absorbed inside normal diurnal variation.

When spreads widen, the difference between "generally correct" and "structurally aligned" becomes visible in the ledger. When portfolios grow, small percentage differences compound. When volatility persists, persistence itself carries informational value. At that point, dispatch logic is no longer a background setting. It becomes part of operational performance.

A battery can follow price. The question is whether it recognizes context.

Technical Paper

The full methodology — including mathematical derivation, multi-year backtests, seasonal behaviour and limitations — is documented in the technical paper.

Download the technical paper (PDF)

The full methodology — including mathematical derivation, multi-year backtests, seasonal behaviour and limitations — is documented in the technical paper.

Download the technical paper (PDF)

Conversation

Control logic compounds. So do small errors.If your storage portfolio still relies on fixed-threshold dispatch, it may be worth quantifying the performance gap. We examine control frameworks against market structure and regime behavior, not historical averages alone.

Contact us.

Posts.

VIEW MORE