Skip to content

Benchmark Dataset

The benchmark dataset is a collection of described biogas plants. It is used to test how well an artificial intelligence (AI) can rebuild a working PyADM1ODE model from the description of a plant.

In a nutshell

An AI reads the description of a biogas plant – as text or as a sketch – and is supposed to rebuild that same plant in the computer. The benchmark measures how accurately it succeeds.

What is this about?

PyADM1ODE is software that can simulate biogas plants. To make the software compute a specific plant, you first have to describe that plant: How many vessels are there? How big are they? How are they connected? Is there a combined heat and power unit?

Exactly this step – from a description in words to a finished plant in the software – is the task the benchmark evaluates. The AI receives a description and produces the plant from it. Its result is then compared against a known reference solution.

flowchart LR
    A["📝 Plant<br>description"] --> B["🤖 AI rebuilds<br>the plant"]
    B --> C["⚖️ Compare with<br>the reference solution"]
    C --> D["📊 Score<br>in percent"]

Why is this useful?

  • Comparability: Different AI models solve the same tasks – you can objectively
    see which one performs better.
  • Realism: The plants are based on real agricultural biogas plants.
  • Understanding: The dataset vividly shows what a biogas plant consists of and
    what matters when modelling one.

Who is this documentation for?

This introduction is aimed at readers without programming experience. No coding knowledge is required – technical terms are explained along the way.

Where to go next