Quick Start

1. Install the most recent version of Julia
2. Install the package EnergyModelsBase and the time package TimeStruct, by running:

julia ] add TimeStruct EnergyModelsBase

These packages are required as we do not only use them internally, but also for building a model.

1. Install the package EnergyModelsGasNetworks

julia ] add EnergyModelsGasNetworks

1. Install the package JuMP by running:

julia ] add JuMP

1. Install a suitable solver

   Some constraints defined in EnergyModelsGasNetworks are nonlinear and nonconvex.

   • Pooling constraints. The pooling formulation introduces bilinear terms (non-convex quadratic), defining a nonconvex QCQP (and, if integer variables are present, a nonconvex MINLP).
   • Pressure-flow constraints for blended gases. The pressure-flow relation for blended gases is also nonlinear and nonconvex (adapted Weymouth equation for blended gases). However, within the physically operating region (i.e., inlet pressure >= outlet pressure, nonnegative flows and bounded pressures), the relation is smooth and monotonic. To obtain a MILP, we approximate the nonlinear surface using a piecewise-affine outer approximation built from tangent planes via PiecewiseAffineApprox.

   Consequently, when pooling and/or pressure-flow constraints are enabled, you need a solver capable of handling nonconvex MINLP problems, such as Alpine, which uses a MIP solver and a continous NLP solver internally.

julia ] add Alpine, Ipopt, HiGHS using JuMP, Alpine, Ipopt, HiGHS ipopt = optimizer_with_attributes(Ipopt.Optimizer, "print_level" => 0) highs = optimizer_with_attributes(HiGHS.Optimizer, "output_flag" => false) model = Model( optimizer_with_attributes( Alpine.Optimizer, "nlp_solver" => ipopt, "mip_solver" => highs, ), )

Using other MIP solvers such as Gurobi or Xpress can significantly improve performance.