Robust Dual Dynamic Programming (RDDP)
Usage
pkg> add RDDPExample
# inventory management under uncertain demand
using RDDP,Gurobi,JuMP
N = 10
msro = RDDP.buildMultiStageRobustModel(
N_stage = N, # number of stages
optimizer = Gurobi.Optimizer, # choose an optimizer
MaxIteration = 100, # maximum epoches
MaxTime = 60, # maximum trainning time
Gap = 0.01, # optimal gap
use_maxmin_solver = false # using a vertex enumeration by default
) do ro::JuMP.Model,t::Int
# defining operator's strategy
@variable(ro,0 <= x <= 5,RDDP.State,initial_value = 0) # state variable (stock level)
@variable(ro,0 <= u <= 1) # control variable (supply)
@variable(ro,0 <= ξ[i = 1:2] <= 1,RDDP.Uncertain) # uncertain parameter (demand)
@constraint(ro,sum(ξ) <= 1) # constraint abount uncertain parameter
@constraint(ro, x.out == x.in - u + sum(ξ)) # state transition (temporal dependence of stock level)
@objective(ro,Min,x) # minimize the stock
end
RDDP.train(msro)Citing RDDP.jl
If you use RDDP.jl, we ask that you please cite the following:
@article{rddp.jl,
title={Enhancing the Flexibility of Storage Integrated Power System by Multi-stage Robust Dispatch},
author={Shi, Yunhui and Guo, Chuangxin and Dong, Shufeng and Chen, Zhe and Wang, Luyu},
journal={IEEE Transactions on Power Systems},
year={2020},
publisher={IEEE}
}
@article{georghiou2019robust,
title={Robust dual dynamic programming},
author={Georghiou, Angelos and Tsoukalas, Angelos and Wiesemann, Wolfram},
journal={Operations Research},
volume={67},
number={3},
pages={813--830},
year={2019},
publisher={INFORMS}
}