sodym

The sodym package provides key functionality for material flow analysis, including

• the class MFASystem acting as a template (parent class) for users to create their own material flow models
• the class NamedDimArray handling mathematical operations between multi-dimensional arrays
• different classes like DynamicStockModel representing stocks accumulation, in- and outflows based on age cohort tracking and lifetime distributions. Those can be integrated in the MFASystem.
• different options for data input and export, as well as visualization

Thanks

sodym is an adaptation of:

ODYM
Copyright (c) 2018 Industrial Ecology
author: Stefan Pauliuk, Uni Freiburg, Germany
https://github.com/IndEcol/ODYM

The development of sodym was conducted within the TRANSIENCE project, grant number 101137606, funded by the European Commission within the Horizon Europe Research and Innovation Programme.

Installation

sodym dependencies are managed with pip.

To install as a user: run python -m pip install sodym@git+https://github.com/pik-piam/sodym.git

To install as a developer:

1. Clone the sodym repository using git.
2. From the project main directory, run pip install -e ".[test,docs,examples]" to obtain all the necessary dependencies, including those for running the tests, making the documentation, and running the examples.

Note that it is advisable to do this within a virtual environment.

Examples

The notebooks in the examples folder provide usage examples of the code.

Why choose sodym?

MFA models mainly consist on mathematical operations on different multi-dimensional arrays.

For example, the generation of different waste types waste might be a 3D-array defined over the dimensions time $t$, region $r$ and waste type $w$, and might be calculated from multiplying end_of_life_products (defined over time, region, and product type $p$) with a waste_share mapping from product type to waste type. In numpy, the according matrix multiplication can be carried out nicely with the einsum function, were an index string indicates the involved dimensions:

waste = np.einsum('trw,pw->trp', end_of_life_products, waste_share)

sodym uses this function under the hood, but wraps it in a data type NamedDimArray, which stores the dimensions of the array and internally manages the dimensions of different arrays involved in mathematical operations.

With this, the above example reduces to

waste[...] = end_of_life_products * waste_share

This gives a sodym-based MFA models the following properties:

• Simplicity: Since dimensions are automatically managed by the user, coding array operations becomes much easier. No knowledge about the einsum function, about the dimensions of each involved array or their order are required.
• Sustainability: When changing the dimensionality of any array in your code, you only have to apply the change once, where the array is defined, instead of adapting every operation involving it. This also allows, for example, to add or remove an entire dimension from your model with minimal effort.
• Versatility: We offer different levels of sodym use: Users can choose to use the standard methods implemented for data read-in, system setup and visualization, or only use only some of the data types like NamedDimArray, and custom methods for the rest.
• Robustness: Through the use of Pydantic, the setup of the system and data read-in are type-checked, highlighting errors early-on.
• Performance: The use of numpy ndarrays ensures low model runtimes compared with dimension matching through pandas dataframes.