Growing shares of renewable generation increase forecast uncertainty and widen the gap between day-ahead market outcomes and the system states that materialize in operation. In the German power system, updated expectations of wind and photovoltaic feed-in trigger intraday trading and schedule adaptations of conventional power plants, substantially changing zonal balances, power flows, and congestion. If these market-driven adjustments are not represented, redispatch assessments may rely on system states that are neither physically nor economically plausible.

This paper presents a market module developed in the joint research project ProMetheUs with German transmission system operators to represent the market response to renewable uncertainty. The contribution is a data-driven methodology that translates renewable realization scenarios into realistic, scenario-dependent adjustments of the conventional generation in Germany and the resulting changes in control-area balances. The approach combines prediction and optimization. First, a multivariate time-series model learns the relation between day-ahead system conditions, realized renewable feed-in patterns, and intraday adjustment behavior. It jointly forecasts balance changes for the German control areas and the redistribution of conventional generation across major fuel types. Second, these aggregated fuel-specific adjustments are translated into plant-level schedules by a mixed-integer linear optimization model with unit commitment decisions. The optimization respects technical constraints such as generation limits, ramp rates, minimum up- and down-times, and unit unavailability. Its objective combines variable generation costs, start-up costs, and plant-specific penalties for deviations from reported schedules. These penalties are calibrated from historical rescheduling behavior so that units with frequent intraday schedule changes are penalized less and are therefore more likely to represent the expected market response.

Model development and evaluation are based on German data from 2022 to 2024, including day-ahead renewable forecasts, realized feed-in, load, control-area balances, planned conventional schedules, and generation availability. A key feature is the distinction between historical validation and operational application. Historical realizations are used for training and ex-post evaluation, while in practice future renewable realization is unknown. For operational use, the trained module is therefore driven by weather-ensemble-based renewable scenarios to generate plausible scenario-specific conventional dispatch adaptations and resulting balance changes. In this way, the method provides realistic, market-consistent system states to support transmission system operators in assessing congestion risks and improving resilience to forecast uncertainty.