Gas Model

All methods for constructing gasmodels should be defined on the following type:

GasModels.AbstractGasModel — Type

root of the gas formulation hierarchy

which utilizes the following (internal) functions:

GasModels.build_ref — Function

Builds the ref dictionary from the data dictionary. Additionally the ref dictionary would contain fields populated by the optional vector of ref_extensions provided as a keyword argument.

source

When using the build_ref for transient problem formulations the following ref extension has to be added to populate the fields necessary for formulate the transient optimization problems.

GasModels.ref_add_transient! — Function

adds the additional data into the ref that is required to used to formulate the transient formulation

source

Network Formulations

Type Hierarchy

We begin with the top of the hierarchy, where we can distinguish between gas flow models. There are currently five formulations supported in GasModels. Two full non convex formulations and three relaxations.

AbstractNLPModel <: AbstractGasModel
AbstractMINLPModel <: AbstractGasModel
AbstractMISOCPModel <: AbstractGasModel
AbstractMIPModel <: AbstractGasModel
AbstractLPModel <: AbstractGasModel

Gas Models

Each of these forms can be used as the type parameter for a GasModel, i.e.:

NLPGasModel <: AbstractNLPForm
MINLPGasModel <: AbstractMINLPModel
MISOCPGasModel <: AbstractMISOCPModel
MIPGasModel <: AbstractMIPModel
LPGasModel <: AbstractLPModel

For details on AbstractGasModel, see the section on Gas Model.

User-Defined Abstractions

The user-defined abstractions begin from a root abstract like the AbstractGasModel abstract type, i.e.

AbstractMyFooModel <: AbstractGasModel

StandardMyFooForm <: AbstractFooModel
FooGasModel = AbstractGasModel{StandardFooForm}

Supported Formulations

All formulation names refer to how underlying physics of a gas network is modeled. For example, the LP model uses a linear representation of natural gas physics. If a model includes valves, then the resulting mathematical optimization problems will be MIP since valve controls are discrete.

Formulation	Steady-State	Transient	Description
NLP	Y	N	Physics is modeled using nonlinear equations.
MINLP	Y	N	Physics is modeled using nonlinear equations. Directionality of flow is modeled using discrete variables
MINLP	Y	N	Physics is modeled using second order cone equations. Directionality of flow is modeled using discrete variables
MINLP	Y	N	Physics is modeled using linear equations. Directionality of flow is modeled using discrete variables
LP	Y	N	Physics is modeled using linear equations.