This information is part of the Modelica Standard Library maintained by the Modelica Association.

Without pre-caution when implementing a medium model, it is very easy that non-linear algebraic systems of equations occur when using the medium model. In this section it is explained how to avoid non-linear systems of equations that result from unnecessary dynamic state selections.

A medium model should be implemented in such a way that a tool is able to select states of a medium in a balance volume statically (during translation). This is only possible if the medium equations are written in a specific way. Otherwise, a tool has to dynamically select states during simulation. Since medium equations are usually non-linear, this means that non-linear algebraic systems of equations would occur in every balance volume.

It is assumed that medium equations in a balance volume are defined in the following way:

  package Medium = Modelica.Media.Interfaces.PartialMedium;
  Medium.BaseProperties medium;
equation
   // mass balance
     der(M)  = port_a.m_flow + port_b.m_flow;
     der(MX) = port_a_mX_flow + port_b_mX_flow;
           M = V*medium.d;
          MX = M*medium.X;

   // Energy balance
   U = M*medium.u;
   der(U) = port_a.H_flow+port_b.H_flow;

Single Substance Media

A medium consisting of a single substance has to define two of "p,T,d,u,h" with stateSelect=StateSelect.prefer if BaseProperties.preferredMediumstates = true and has to provide the other three variables as function of these states. This results in:

• static state selection (no dynamic choices).
• a linear system of equations in the two state derivatives.

Example for a single substance medium

p, T are preferred states (i.e., StateSelect.prefer is set) and there are three equations written in the form:

d = fd(p,T)
u = fu(p,T)
h = fh(p,T)

Index reduction leads to the equations:

der(M) = V*der(d)
der(U) = der(M)*u + M*der(u)
der(d) = der(fd,p)*der(p) + der(fd,T)*der(T)
der(u) = der(fu,p)*der(p) + der(fu,T)*der(T)

Note, that der(y,x) is the partial derivative of y with respect to x and that this operator is available in Modelica only for declaring partial derivative functions, see Section 12.7.2 (Partial Derivatives of Functions) of the Modelica 3.4 specification.

The above equations imply, that if p,T are provided from the integrator as states, all functions, such as fd(p,T) or der(fd,p) can be evaluated as function of the states. The overall system results in a linear system of equations in der(p) and der(T) after eliminating der(M), der(U), der(d), der(u) via tearing.

Counter Example for a single substance medium

An ideal gas with one substance is written in the form

redeclare model extends BaseProperties(
   T(stateSelect=if preferredMediumStates then StateSelect.prefer else StateSelect.default),
   p(stateSelect=if preferredMediumStates then StateSelect.prefer else StateSelect.default)
equation
   h = h(T);
     u = h - R_s*T;
     p = d*R_s*T;
    ...
end BaseProperties;

If p, T are preferred states, these equations are not written in the recommended form, because d is not a function of p and T. If p,T would be states, it would be necessary to solve for the density:

   d = p/(R_s*T)

If T or R_s are zero, this results in a division by zero. A tool does not know that R_s or T cannot become zero. Therefore, a tool must assume that p, T cannot always be selected as states and has to either use another static state selection or use dynamic state selection. The only other choice for static state selection is d,T, because h,u,p are given as functions of d,T. However, as potential states only variables appearing differentiated and variables declared with StateSelect.prefer or StateSelect.always are used. Since "d" does not appear differentiated and has StateSelect.default, it cannot be selected as a state. As a result, the tool has to select states dynamically during simulation. Since the equations above are non-linear and they are utilized in the dynamic state selection, a non-linear system of equations is present in every balance volume.

To summarize, for single substance ideal gas media there are the following two possibilities to get static state selection and linear systems of equations:

1. Use p,T as preferred states and write the equation for d in the form: d = p/(T*R_s)
2. Use d,T as preferred states and write the equation for p in the form: p = d*T*R_s

All other settings (other/no preferred states etc.) lead to dynamic state selection and non-linear systems of equations for a balance volume.

Multiple Substance Media

A medium consisting of multiple substance has to define two of "p,T,d,u,h" as well as the mass fractions Xi with stateSelect=StateSelect.prefer (if BaseProperties.preferredMediumStates = true) and has to provide the other three variables as functions of these states. Only then, static selection is possible for a tool.

Example for a multiple substance medium:

p, T and Xi are defined as preferred states and the equations are written in the form:

d = fp(p,T,Xi);
u = fu(p,T,Xi);
h = fh(p,T,Xi);

Since the balance equations are written in the form:

  M = V*medium.d;
MXi = M*medium.Xi;

The variables M and MXi appearing differentiated in the balance equations are provided as functions of d and Xi and since d is given as a function of p, T and Xi, it is possible to compute M and MXi directly from the desired states. This means that static state selection is possible.