Load Estimation in Onshore Wind Farms Using Surrogate Modelling and Generic Turbine Models

Load Estimation in Onshore Wind Farms Using Surrogate Modelling and Generic Turbine Models This record provides all necessary materials to reproduce the analyses and results presented in our paper, currently under review, on wind farm load estimation. The codebase — including data processing pipelines, analysis scripts, and visualization tools — is maintained separately on GitHub (see link below). Dependencies Python version: 3.11.1 This project uses a local copy of OpenOA v.2.3 (https://github.com/NREL/OpenOA). Other dependencies are listed in the requirements.txt Usage To reproduce the results, follow these steps: Download the code repository (see link below). Download the dataset from Zenodo. Unzip data.zip and replace the existing data/ directory in the code repository. Install the required dependencies. The analysis workflow is organized in Jupyter notebooks: 01_proc_raw_data.ipynb: Initial data processing and cleaning 02_sampling.ipynb: Data sampling procedures 03_gpr_training.ipynb: Gaussian Process Regression model training 03_pce_training.ipynb: Polynomial Chaos Expansion model training 04_case_study.ipynb: Application and evaluation of models in case study Data Structure data/ case_study/ predictions.pkl: Stored model predictions for the case study analysis models/ gpr_models.pickle: Trained Gaussian Process Regression models pce_models.pickle: Trained Polynomial Chaos Expansion models samples/ sample_set.npy: Output of the 02_sampling.ipynb notebook. scada/ farmdata_*: Processed farmdata including all turbines, ready for application simulation/ sample_sim_setup/: A set of representative openfast files for one simulation case. casematrix.csv: Simulation case definitions. Transformed to .csv from `sample_set.npy` surrogate_data.csv: Processed 10min load variables by case number turbines/ IEA-3.4-130-RWT/: IEA reference wind turbine model files (https://github.com/IEAWindSystems/IEA-3.4-130-RWT) Adapted RWT model/: Above model with minor changes to better fit case study. Contact Alexander Mönnig: [e-mail] Ulrich Römer: [e-mail]