Pre-Release of FMI Layered Standard for Differential-Algebraic Equations (FMI-LS-DAE) v1.0.0-alpha.1

The FMI Project is happy to announce the first alpha pre-release of the FMI Layered Standard for Differential-Algebraic Equations (FMI-LS-DAE), which allows FMI 3.0 Model Exchange FMUs to expose their model equations as a semi-explicit index-1 system of differential-algebraic equations (DAE) instead of an ordinary differential equation (ODE) system.

Many thanks to the whole working group and especially to the working group leaders Andreas Heuermann (Santa Anna Research Institute) and Joel Andersson (FMIOPT)!

Summary: In addition to the continuous states of FMI 3.0 Model Exchange, this layered standard introduces algebraic variables and the corresponding residual equations, which are solved by the importer instead of inside the FMU. The additional information is described in the layered standard manifest file, which is part of the FMU archive, so that no extension of the FMI 3.0 C API is needed. A structural parameter declared by the FMU lets the importer switch between the classical ODE mode and the new DAE mode, which keeps such FMUs usable by importers that do not support this layered standard.

Key features:

• A layered standard manifest stored inside the FMU at /extra/org.fmi-standard.fmi-ls-dae/fmi-ls-manifest.xml, defined by the XML schema fmi3LayeredStandardDaeManifest.xsd.
• Declaration of the algebraic variables and of the residual equations of the DAE in the manifest, together with the model structure and the dependencies of outputs, continuous state derivatives, event indicators and residuals. The dependency information preserves the sparsity of the DAE system, which is lost in the corresponding reduced ODE system.
• A structural parameter enableDAEModeParameter that switches the FMU between ODE mode and DAE mode. An FMU implementing this layered standard is at the same time a valid FMI 3.0 Model Exchange ODE FMU and defaults to ODE mode, so that importers that do not support this layered standard can use it unchanged.
• No extension of the FMI 3.0 C API is required. Algebraic variables and residuals are exchanged through the existing FMI 3.0 getter and setter functions.

Exporting a model in DAE form addresses a number of long-standing issues:

• Models can be exported without internal index reduction, which can improve accuracy through better handling of drift.
• Local nonlinear solvers inside the FMU become unnecessary, avoiding mismatches between local and global error tolerances.
• The sparsity of the original equation system is preserved instead of being lost during the reduction to an ODE, which allows importers to exploit sparsity for better performance.
• The cost of the symbolic DAE-to-ODE transformation is avoided, which is particularly relevant for very large models, such as simulation models representing power grids on a national or larger scale.

For the full picture please check out the preprint of our paper “Towards an FMI Layered Standard for DAE: Applications for Simulation and Optimization” for the coming American Modelica and FMI Conference 2026 on arXiv: https://arxiv.org/abs/2606.22544 It also reports successful tests in the following tools: CasADi, FMIOPT, Simcenter Twin Activate, and MOO (the dynamic optimization tool of OpenModelica).

As an alpha pre-release, the specification is published for review and for prototype implementations, and it is still subject to change before the v1.0.0 release. Known limitations of this version are the restriction to semi-explicit index-1 DAEs, the consistent initialization in DAE mode, and the handling of discontinuities and events, which are not specified yet. Feedback from tool vendors and users is very welcome and is best provided as an issue in the FMI-LS-DAE GitHub repository.

You can inspect the specification of this pre-release here: https://modelica.github.io/fmi-ls-dae/v1.0.0-alpha.1/ and the current development version here: https://modelica.github.io/fmi-ls-dae/main/

For the general concept of Layered Standards to the FMI Standards see this paper: https://doi.org/10.3384/ecp204381

Learn more on the Release page on Github.