Further parameters

There are two more structs that are relevant for the user when running loglinearSDDP.

ProblemParams

struct ProblemParams
    number_of_stages::Int64
    number_of_realizations::Int64
    tree_seed::Union{Nothing,Int}
    gurobi_coupling_index_start::Union{Nothing,Int}
    gurobi_cut_index_start::Union{Nothing,Int}
    gurobi_fix_start::Union{Nothing,Int}
end

This struct is used to store some parameters of the test problem that is solved. This is mainly used for logging purposes.

Its fields are defined as follows:

• number_of_stages: number of stages of the multistage problem
• numer_of_realizations: number of realizations of the stagewise independent noise $\eta$
• tree_seed: seed for the scenario tree generation
• gurobi parameters: these parameters refer to Gurobi-internal indices of coupling constraints, cut constraints and cut intercept variables; if they are known, they can be provided by the user to speed up parts of the computations (see more details below)

Note that the size of eta in AutoregressiveProcessStage should match number_of_realizations defined in ProblemParams.

AlgoParams

using LogLinearSDDP, SDDP

mutable struct AlgoParams
    stopping_rules::Vector{SDDP.AbstractStoppingRule}
    simulation_regime::LogLinearSDDP.AbstractSimulationRegime
    cut_selection::Bool
    print_level::Int64
    log_frequency::Int64
    log_file::String
    run_numerical_stability_report::Bool
    numerical_focus::Bool
    silent::Bool
    forward_pass_seed::Union{Nothing,Int}
    run_description::String
    model_approach::Symbol
end

This struct is used to store some parameters that control the SDDP algorithm that is used to solve the model. Many of these parameters are the same that the train function in SDDP.jl uses as arguments.

Its fields are defined as follows:

• stopping_rules: A vector of different SDDP stopping rules (same as for SDDP.jl).
• simulation_regime: Defines which type of simulation should be used for the in-sample simulations (see simulations).
• cut_selection: Controls if a cut selection scheme should be applied. Note that this is not supported yet but might be added in the future, so we added a placeholder.
• print_level, log_frequency, log_file: As for train in SDDP.jl, these parameters control the logging of the SDDP output
• run_numerical_stability_report: As for train in SDDP.jl, this parameter controls the generation of numerical stability reports. Note that this is only a placeholder so far and should always be set to False.
• numerical_focus: If set to True, Gurobi will set its numerical_focus parameter to True, which may help in case of numerical issues but will also slow down the solver.
• silent: Controls if the solver output is printed on the console.
• forward_pass_seed: Seed that is used for sampling realizations of the stagewise independent noise $\eta$ in the forward pass of SDDP.
• run_description: Allows the user to add some description to a specific model run that is used in the log-file.
• model_approach: For logging purposes, stores which concrete model is used for the stochastic process (i.e. LOG-BIC, LOG-1, LIN-FIT or LIN-SHA for our hydrothermal scheduling problem, see experimental set-up).

This page was generated using Literate.jl.