Compute Optimal Investment Candidates for H2 Systems — MultiStock_H2_Investment_reward_compute

Searches for the optimal investment solution among bounded candidates for H2 systems, including storage, must-run clusters, and flexible clusters.

Usage

MultiStock_H2_Investment_reward_compute_once(
  areas_invest,
  max_ite,
  list_storage_bounds,
  storage_points_nb,
  candidates_types_gen,
  penalty_low = 5000,
  penalty_high = 5000,
  penalty_final_level = 5000,
  opts,
  mc_years_optim,
  cvar = 1,
  storage_annual_cost,
  nb_sims = 51,
  parallelprocess = F,
  nb_sockets = 0,
  unspil_cost = 3000,
  file_intermediate_results,
  list_ratio_max_hydro,
  remove_cluster = F
)

Arguments

areas_invest: `character vector` Names of the areas/zones to optimize.
max_ite: `integer` Maximum number of iterations per area.
list_storage_bounds: `list` List of vectors of two integers (min, max) for each area, specifying the storage bounds to test.
storage_points_nb: `integer` Number of storage points to test at each iteration (must be > 3 for effective refinement).
candidates_types_gen: `data.frame` A data.frame describing the generation candidates to be optimized. **See section ‘Structure of candidates_types_gen’ for details about required columns and their usage.**
penalty_low: `numeric` Penalty when storage is below the guide curve (default: 5000).
penalty_high: `numeric` Penalty when storage exceeds the guide curve (default: 5000).
penalty_final_level: `numeric` Penalty for out-of-bounds final storage level (default: 5000).
opts: `list` Study parameters (typically the output of `antaresRead::setSimulationPath(simulation = "input")`).
mc_years_optim: `integer vector` Monte Carlo years used for optimization.
cvar: `numeric [0,1]` Probability used in `Grid_Matrix()` (default: 1).
storage_annual_cost: `numeric` Annual storage cost in €/MWh.
nb_sims: `integer` Number of simulations per evaluation (default: 51).
parallelprocess: `logical` Whether to use parallel processing for Water values calculations.
nb_sockets: `integer` Number of sockets to use if `parallelprocess = TRUE`.
unspil_cost: `numeric` Unspilled energy cost in €/MW for all relevant areas.
file_intermediate_results: `character` Local path for saving intermediate results for backup/restart.
list_ratio_max_hydro: `list` For each area, gives the ratio used to deduce maximum weekly power for the dimensioning (used at weekly resolution only).
remove_cluster: `logical` Should candidate clusters be removed before running the investment process.

Value

A list, with each element corresponding to an optimized area containing:

`all_costs`: Complete grid of total costs for all tested candidates,
`best`: The best candidate(s) found,
`last_storage_points`, `last_candidates_data`: Bounds and parameters used in the final iteration,
`optim_time`: Optimization duration,
`reward`: Reward function used,
`optimal_traj`: Optimal storage trajectory,
`optimal_max_hydro`, `watervalues`: Optimal hydraulic parameters, etc.

Details

The function first computes gain (reward) functions using the 'plaia' methodology (via Monte Carlo simulations). For each area, it then iterates over all candidate combinations, calculating the operational cost for each using dynamic programming (`Grid_Matrix()`). The operational cost computation is based on optimal use (value functions) that are themselves modified by each candidate's characteristics (storage size, cluster types, etc.). At each iteration, the tested domain is refined around the best candidate solution from the previous iteration (grid refinement).

The function optimizes the **installed power** (e.g. turbine/pump/generator capacities) of candidates, but only to determine the **maximum weekly energy or power** that can be delivered or absorbed. All constraints and optimizations are performed on a weekly basis. **Hourly power-related congestion or inadequacies cannot be detected**: only weekly energy or power limitations are modeled in the optimization. This means that the function is suited for weekly energy sizing, but will not capture operational issues that arise at an hourly timescale.

Note

Power sizing is optimized, but for weekly assessment only. If you need to capture possible hourly power congestion or short-term operating constraints, you will need a different or more detailed optimization.

Structure of candidates_types_gen

The candidates_types_gen argument must be a data.frame with at least the following columns:

index: Integer. Unique identifier for each candidate (should be 1, 2, ..., n).
name: Character. Name of the candidate (used for study and result traceability).
type: Character. Type of cluster: either `"cluster_flexible"` (dispatchable/flexible) or `"cluster_bande"` (must-run or base load).
TOTEX: Numeric. Total annualized cost per MW of the candidate (euros/MW/year).
Marg_price: Numeric. Marginal cost for the candidate (euros/MWh).
Part_fixe: Numeric, between 0 and 1. Fixed part of a flexible cluster (only for `type == "cluster_flexible"`). Value of 0 if not applicable.
Prix_fixe: Numeric. Fixed price (euros/MWh) for the fixed part of a flexible cluster.
Borne_min: Numeric. Minimum possible size (MW) for the candidate.
Borne_max: Numeric. Maximum possible size (MW) for the candidate.
Points_nb: Integer (\>=4). Number of tested candidate positions (grid points) for this cluster per iteration.
Zone: Character. Name of the area/zone where this candidate can be added.

Notes

Flexible clusters are automatically split into fixed and variable parts if needed.
Grid ranges are updated at each iteration based on the best found solution (refinement approach).
The function manages Monte Carlo simulations and ensures reproducibility via the intermediate result file.
Only weekly (not hourly) congestion or adequacy issues can be assessed.