Supplementary data: "Revealing drivers and risks for power grid frequency stability with explainable AI"

This repository contains processed data and result files for the paper Revealing drivers and risks for power grid frequency stability with explainable AI.  The code for producing the processed data and the results is available at github.

Data

For each area, the data folder contains the feature and target data used to train the ML model.

• raw_input_data.h5 : The aggregated external features without additional engineered features.
• input_forecast.h5 and input_actual.h5: The day-ahead available (forecast) and ex-post available (actual) data of external features including the engineered features.
• outputs.h5 :  The grid frequency stability indicators.
• version_2021-07-01: Folder containing the training and test sets used for the results.
• documentation_of_data_download: Plots and information files concerning the ENTSO-E raw data and its aggregation.

Data sources

The data for input features (raw_input_data.h5, input_forecast.h5 and input_actual.h5) is derived from ENTSO-E Transparency Platform data [1]. The target data (outputs.h5) is based on power grid frequency recordings from the German Transmission System Operator TransnetBW [2].

Results

For each area and each target, the result folder comprises the results of hyper-parameter optimization, model prediction and interpretation via SHAP. The results refer to the full model and the restricted model (containing only day-ahead features).

• cv_results_gtb_full.csv and cv_results_gtb_day_ahead.csv : Performance results for each combination in the hyper-parameter grid search.
• cv_best_params_gtb_full.csv and cv_best_params_gtb_day_ahead.csv : Hyper-parameters used in the final (optimized) model.
• shap_values_gtb_full.npy and shap_interaction_values_gtb_full.npy : First-order SHAP values and second-order SHAP interaction values for the full model.
• y_pred.h5 : Predictions of daily profile predictor, full model and day-ahead model.

Disclaimer

The data might be subject to copyright or related rights. Please consult the primary data owner.