wtuq (Wind Turbine Uncertainty Quantification framework)

Purpose

This Uncertainty Quantification (UQ) framework is a wrapper around the open-source uncertainpy package. It provides extensions and interfaces especially useful for the implementation of UQ in aeroelastic wind turbine simulations.

Structure

The framework consists of two main parts. wtuq_framework is an installable package which is independent of any wind turbine or research question specific input. The modules in this package can rather be seen as direct extensions to the uncertainpy package and might in fact (in part) be merged into the uncertainpy package in the future. The use_cases directory contains wind turbine and research question specific routines. For now, only one use case has been implemented.

- wtuq_framework -> installable package
  | - uq_framework.py
  | - splining.py
  | - helperfunctions.py
  | - UQResultsAnalysis.py
- use_cases
  | - stability_analysis
    | - config
    | - reference_data
    | - result_comparison
    | - tool_interfaces
    | - run_analysis.py
    | - preprocessor.py
    | - postprocessor.py
    | - etc etc.py
  | - other use cases
- doc
- tests

Usage

1. Make a config for the framework run

   The config has two parts. The [framework] part provides the settings for uncertainpy and the definition of the uncertain parameters. The specification of this part is given in wtuq_framework/config.spec. The [use_case] part contains use_case specific information. A new file of config specifications has to be generated for new use cases. An example can be found in use_cases/stability_analysis/config/use_case_config.spec

   (For an example config file, see use_cases/config/examples/example_config_case_study1.txt)

2. Run the framework

   Basic use: python run_analysis.py -c path_to_your_config

   There are two alternative 'restart' functionalities.

   1. python run_analysis.py -c path_to_your_config -rh5 ../reference_data/bladed/run123/uq_results/Model.h5

      The framework can extract the training data of a previous run from the standard uncertainpy .h5 output file. The script will check that the requested Model.h5 file is available and that it has the same uncertain parameter settings as requested in the config file.

      This setting is useful when changes are made to uncertainpy itself or if new postprocessing scripts need to be tested.

   2. python run_analysis.py -c path_to_your_config -r ../reference_data/bladed/run123/*

      The framework also has a restart capability that recognizes already computed results and omits the corresponding simulations. This way, different post-processing algorithms and settings can be applied and compared. It also helps to recover an analysis in case of a software crash.

      The restart directory are defined as a pattern (as understood by the glob package) and must be given with a full valid path.

3. Compare different runs

   The UQ results of different runs (e.g., different models/tools) can be compared by running the wtuq_framework.uq_results_analysis script with a special config file.

   Basic use: python uq_results_analysis.py -c path_to_your_uq_results_analysis_config

   The config specification can be found in wtuq_framework/uq_results_analysis_config.spec. An example config file can be found in use_cases/result_comparison/config_torque_paper.txt

Requirements

General

• OS: Windows & Linux
• Python >=3.7

Required Python packages

• numpy
• matplotlib
• pandas
• configobj
• bokeh
• uncertainpy @ git+https://github.com/DLR-AE/uncertainpy.git
• pydmd
• geomdl

Installing

• to use: pip install --user https+git://github.com/DLR-AE/wtuq.git
• to develop
  • clone repository
  • pip install -e <local_repo_path>

Only the general wtuq_framework package will be installed automatically. To work on the available use cases, the repository has to be cloned.

Support

Please do not hesitate to contact us if you have any questions on the framework or if you want more information. We are open for collaboration and welcome further development or modification of the framework. You can contact us on [email protected] / [email protected], or by reaching out to any of the other collaborators.

Contributing

This framework was the joint work of the DLR (Institute of Aeroelasticity), Leibniz University Hannover (Institut für Windenergiesysteme), and Nordex GmbH. Active collaborators on the project were:

• Hendrik Verdonck, [email protected]
• Oliver Hach, [email protected]
• Claudio Balzani, [email protected]
• Jelmer Polman, [email protected]
• Sarah Müller, [email protected]
• Johannes Rieke, [email protected]
• Otto Braun

Acknowledgment

This software package was developed in the frame of the German national research project QuexUS. This project is funded by the German Federal Ministry for Economic Affairs and Climate Action, grant no. 03EE3011A/B.