Supplementary data: "Physics-inspired machine learning for power grid frequency modelling"

This repository contains processed data and result files for the paper Physics-inspired machine learning for power grid frequency modelling . The code for producing the processed data and the results is available at github. <strong>Data</strong> The data folder contains the feature data used to train the machine learning model. The target data, i.e., the grid frequency measurements, were taken from this repository and are not available in here. The data for Continental Europe comprises the following folders and files: <em>raw_input_data.h5</em><strong> </strong>:<strong> </strong>The aggregated external features without additional engineered features. <em>inputs_forecast.h5 </em><em>:</em> The day-ahead available input features including the engineered features. <em>inputs_actual.h5 </em><em>:</em> Actual (ex-post available) input features including the engineered features. <em>documentation_of_data_download</em>: Information files concerning the ENTSO-E raw data and its aggregation. The data for input features (<em>raw_input_data.h5</em> and <em>input_&lt;type&gt;.h5</em>) is derived from ENTSO-E Transparency Platform data. <strong>Results</strong> The result folder comprises the results of hyper-parameter optimisation, scaling variation and interpretation via SHAP. In particular, it contains these sub-folders and files: <em>tuning </em>: Results of hyper-parameter tuning. <em>best_model </em>: Weights of the trained model with best hyper-parameters. <em>best_model_&lt;scaling-variation&gt; </em>: Weights of the trained models with best hyper-parameters but with a variation of the parameter scaling. <em>fixed_model_hps.pkl </em>: Hyper-parameters that are not optimised. <em>shap_values_&lt;parameter&gt;.h5</em> : SHAP values for the prediction of the system parameters. <strong>Disclaimer</strong> The data might be subject to copyright or related rights. Please consult the primary data owner.