Polish Dispatch Model

How the model optimises battery dispatch across Polish markets

Getting Started

The Polish dispatch model builds on the core dispatch model framework. For an overview of the dispatch model architecture and optimisation approach, see the Core Dispatch Model documentation.

For an overview of the available revenue streams and market structures specific to Poland, see the Revenue Stack page.

Simulated revenues rather than measured

The Poland benchmark simulates the dispatch of a representative battery against public market data. Poland has no asset-level disclosure regime for utility-scale storage: PSE publishes system-wide balancing prices and volumes, but not the schedules or settlement of individual batteries. Revenues are therefore modelled, not reconstructed from metered operations.

The representative asset is a 50 MW standalone battery, modelled at both 2-hour and 4-hour durations. Co-located solar and standalone solar are not supported in this release.

Two-step Optimisation

Poland uses a simpler wholesale structure than Germany, with no intraday gate. The model solves in two sequential steps:

1. Day-ahead step – co-optimises day-ahead energy, FCR capacity, and aFRR capacity with foresight across the delivery day. Commitments are fixed at the end of this step.
2. aFRR energy step – from the PICASSO go-live date, dispatches activated aFRR energy in a rolling real-time window, honouring the day-ahead commitments.

There is no intraday repositioning stage. Day-ahead commitments are set before the aFRR energy step runs.

While Poland’s day-ahead market settles in 15-minute intervals, the day-ahead optimisation step runs at hourly resolution to keep solve times manageable. This is a modelling simplification for this release, not a statement about market design; a future version may move to full 15-minute co-optimisation.

Reserve capability is limited by the battery’s operating point

Poland applies a distinctive rule to how much reserve a battery can offer. Reserve capability is assessed from the battery’s current scheduled output — its operating setpoint — not from how much energy is in the pack. This is a regulatory artefact of a framework originally designed for conventional and hydro plants, and it is restrictive for batteries.

A battery sitting at half charge is not free to offer full reserve in both directions simply because it has energy and headroom available. The market asks where the unit is running right now, not how much energy it is holding. For a unit at a given operating point:

• Upward reserve is limited by the room to increase discharge or reduce charging.
• Downward reserve is limited by the room to increase charging or reduce discharge.
• The two directions cannot both be offered at full band, because they share the same inverter rating measured against the operating point.
• A battery sitting idle at zero output must choose a direction; it cannot hold reserve in both directions while doing nothing.

The table below shows how this works for a 20 MW battery:

This matters because a model that only checked energy headroom would let a battery stack full upward and full downward reserve from a mid-charge position, materially overstating Polish reserve revenue. Tying the offer to the operating point prevents this. Within that limit, upward and downward reserve remain separate products, each priced independently.

Central dispatch and balancing

Poland operates a central dispatch model. PSE runs an integrated scheduling process that decides when units charge and discharge, selecting battery offers in merit order. Since June 2024, imbalances settle at a single price per settlement period, regardless of direction. Passive imbalance income is not monetised in this release.

FCR modelling

• Asymmetric and capacity-only – upward and downward FCR are independent products, each paid its own clearing price. No energy payment is attached to FCR, and no derating is applied to the offered volume.
• Hourly procurement – capacity is contracted in hourly blocks rather than the four-hour blocks used in Germany.
• Limited-energy rule – the battery keeps a short buffer of stored energy in each direction so it can sustain its committed response.

aFRR modelling

• Capacity – priced at the marginal clearing price published by PSE, with upward and downward volumes offered independently.
• Energy activation – from the PICASSO go-live date, activated aFRR energy flows through the battery’s normal charge and discharge, so both revenue and the effect on state of charge are captured directly. Activation in each period is capped by the typical historical activation level for that time of day.

Forecast reserve saturation

In the forecast, reserve revenues decline as the modelled battery fleet grows and competition for reserve increases. Capacity payments for FCR and aFRR saturate over the forecast horizon in line with the projected buildout. The aFRR energy activation cap is scaled by the same saturation signal, so activated-energy revenue falls in step with the capacity markets rather than staying frozen at historical levels. This keeps the reserve stack internally consistent and prevents activated energy from becoming an implausibly large share of revenue in later years.

Revenue calibration

An 80% calibration factor is applied to modelled backtest revenues, consistent with the approach used for the Germany and Great Britain benchmarks. It accounts for the gap between an optimised model with foresight and real-world operation, where forecasting error, availability, and execution reduce captured value.

The factor is applied only to backtest revenues. Forecast revenues are left uncalibrated until a forecast-specific factor is derived. Because the calibration is already applied within the model, backtest figures should not have it applied a second time downstream.

Related Pages

• Revenue Stack – Overview of available revenue streams in Poland
• Core Dispatch Model – Global dispatch model architecture and methodology