MCP Modifying States

[HansOlsson]

The idea is to support updating model components with states based on more accurate measurements of those states during the simulation, and have that as part of the model (and not as some tool-specific solution).

This is just in the preparation stage for making an MCP. There will be an additional presentation at the Dresden meeting (at least that is planned).

The basic idea is presented in section 4.2 of:
https://modelica.org/events/modelica2017/proceedings/html/submissions/ecp17132517_OlssonMattssonOtterPfeifferBurgerHenriksson.pdf

[HansOlsson]

Presented.

Alternatives:

• stateSelect.Measurement and binding equation
• replace “reinit” by “measurement”-operator, or forceState(…) or possibly reinit within a Clock – check if possible
• annotations does not seem like a good idea

Henrik: Consider a proper observer-based variant (instead of just measurement-input), how would that differ?

To clarify: The "measurement" can be a pure measurement, a filtered measurement, or constructed using an observer based on the actual measurement-input, but in the latter case it might be that we should/could combine the observer recreating the state with the feedback linearization in some way.

[HansOlsson]

Using reinit does not work straight away for modifying states in this way.

There are maybe some additional minor detail(s) - but the main reason is straightforward:

We want to replace the state-value with the measurement value inside the equations (and deselect it as a state) in particular in the current evaluation, whereas reinit is done after the model evaluation according to #578

This also implies that the equations for modifying states will be part of the sorting - whereas current reinit doesn't influence sorting.

We could say that reinit behaves differently depending on whether it is in a discrete when-clause or not, but it may be clearer to have another operator.

[HansOlsson]

Now an MCP: https://github.com/modelica/ModelicaSpecification/tree/MCP/0036/RationaleMCP/0036