Processed Distribution-Level Voltage Dataset for Forecasting and State Classification

This dataset contains real-world, long-term voltage measurement data collected from multiple transformer substations within a medium-voltage distribution network in T\u00fcrkiye. The data were obtained from an operational supervisory control and data acquisition (SCADA) system and consist of hourly voltage measurements aggregated at the transformer level over several consecutive years. To ensure confidentiality, all company identifiers, geographic coordinates, and proprietary SCADA IDs have been removed, and the dataset has been fully anonymized.The dataset provides processed voltage-related features, including per-unit normalized mean and minimum voltage values, along with derived indicators for voltage sag events based on IEEE 1159 power quality definitions. In addition, time-aligned labels and statistical features suitable for time-series forecasting and voltage state classification tasks are included. These features enable the evaluation of short-term voltage forecasting models as well as the classification of voltage operating states such as normal conditions, voltage fluctuations, and voltage sags.The dataset is intended to support research on power quality analysis, data-driven voltage forecasting, and intelligent monitoring of distribution systems using machine learning and deep learning techniques. It can be directly used for benchmarking predictive models, classification algorithms, and hybrid forecasting\u2013classification frameworks in smart grid applications.