Skip to content

Assumptions on the problem

Breakdown of weekly supproblems

  • All uncertainties known within the week are considered, with no visibility on other weeks (except for long-term storage through the heuristic’s deliberately limited visibility, and probabilistically for VU).​

  • Over-optimization of dispatch → lower marginal prices.​

  • Dynamic constraints between weeks are taken into account through cyclicity, which assumes that the residual consumption is the same at the beginning and at the end of the week.​

Centralized planner with 1-hour time steps​

  • Simple market bid formats.​

  • Actor strategies inconsistent with the minimization of total system costs (EVs not exposed to market prices, monopoly situations leading actors to influence prices to maximize their profits).​

  • Actors arbitrate between preserving flexibility to minimize imbalance settlement, or conversely not necessarily remaining balanced.​

  • More successive optimizations are required to converge toward the minimum cost so that all actors have the same information.​

DC assumptions on the network

  • The studied network is a high-voltage grid: the reactance \(X\) is much greater than the resistance \(R\) in a high-voltage network, therefore \(X \gg R\).​

  • The network is highly meshed with relatively short line lengths: the voltage angle difference between two nodes is small.​

  • The voltage level \(V\) is similar along the same line between two nodes: the network is meshed with sufficiently well-distributed generation and consumption.​

Aggregation of zones

  • The zone is assimilated to a copper plate with optimal dispatch.​

  • In NTC, the influence on lines is independent of the location of injections/withdrawals within the zone. In FB and in equivalent network models, part of this information is recovered.​

Representatino of line capactities

  • Maximum/minimum flows obtained from different studied supply demand equilibrium situations on the detailed network (in \(N\) or \(N-1\)) → this method overestimates/underestimates capacity.​

  • Linear regression between the equivalent branch flow (sum of line flows) and the maximum congestion rates of lines (in \(N-1\) or \(N\)) → this method is a compromise.​

Aggregation of generation assets

  • Clusters include units with the same characteristics.​

  • Storages have the same values of P_inj / stock_max and P_inj / P_withdraw, otherwise flexibility would change.​

Simplified modeling of assets

  • Incomplete technical constraints (e.g., two deep ramp-downs for nuclear units).​

Simplified optimization

  • Prices are determined with fixed UC (Unit Commitment); the impact of a start-up is not taken into account → lower marginal prices.​

  • Negative or very low prices require the integration of curtailable power, restrictions on pumping/turbining, and the inclusion of start-up/shutdown information for units (MILP).​

  • Solution close to the optimum but not exact.​