Prediction of Total Power Load in Austria

Context and Methodology

This dataset was created as part of a machine learning project in the domain of energy forecasting and smart grid optimization.

The goal of the project is to predict Austria’s total electrical load (MW) at 15-minute intervals based on historical load data.

The primary purpose of the dataset is to train, validate, and evaluate machine learning models for short-term load forecasting, which is critical for grid stability, energy trading, and renewable energy integration.

The dataset was derived from real-world operational data provided by Austrian Power Grid (APG), containing historical total load measurements recorded every 15 minutes starting from January 1, 2023.

Preprocessing steps included handling missing timestamps, cleaning formatting inconsistencies, and engineering basic time-based features.

Technical Details

The dataset is structured as follows:

• Processed data: Subsets for training, validation, and testing, stored separately and referenced by persistent identifiers (PIDs) in DBRepo.

• Output: A trained machine learning model file (.pkl).

Further Details

Users intending to reuse the dataset should be aware of:

• The temporal resolution: 15 minutes between measurements

• Time zone: Central European Time (CET/CEST), including daylight saving shifts

• Potential missing or interpolated data points around daylight saving changes (not unusual in energy datasets)

• Proper preprocessing (feature engineering) steps that must be replicated for correct model use

Detailed README files accompany both the dataset and the codebase, explaining preprocessing logic, feature extraction methods, and model usage instructions to ensure full reproducibility.