Electrical consumption and demand Dataset from the Federal Institute of Espírito Santo, Serra, Brazil

Published: 24 September 2026| Version 2 | DOI: 10.17632/c9sb8cyn86.2
Contributors:
,
,
,
,
,

Description

The "Electrical consumption and demand Dataset from the Federal Institute of Espírito Santo, Serra, Brazil" provides real-world telemetry measurements of electricity consumption and demand from a public campus integrated with a 119 kWp photovoltaic system. Spanning 45 months from October 2022 to July 2026, the dataset comprises over 116,000 records sampled at 15-minute intervals. Measurements were collected using a bidirectional ISSO three-phase power analyzer (DMI F3000R v2 True RMS) installed at the main low-voltage switchboard (QGBT) of the IFES Serra campus. Recorded parameters include voltage, current, frequency, apparent, reactive, and active (direct/reverse) power, power factors, and phase unbalance metrics. The dataset is hierarchically structured into raw and processed folders. The raw data directory (raw_data) contains unedited CSV files (with a .xls extension) featuring original Portuguese headers organized in yearly subdirectories. The processed directory (processed_data) offers consolidated Excel files (.xlsx) with English-translated headers, timestamps synchronized to local time (UTC-3) and UTC, active power scaled to kW, and cumulative active energy computed in kWh for direct consumption and reverse PV injection. An automated four-step quality validation protocol assessed the data: chronological consistency checks identified 80 time-series gaps from power or telemetry outages; ANEEL PRODIST signal quality filtering flagged 346 voltage readings outside the 117–133 V threshold; power-balance checks confirmed phase-summation consistency within 0.1 W/VA; and Ohm's law cross-validation detected only two entries with relative error exceeding 1%. Limitations include aggregated switchboard-level metering without individual load disaggregation, a 15-minute interval that misses fast transients, lack of meteorological data, and load patterns specific to an educational institution. Overall, the dataset provides authentic empirical measurements capturing real-world operational and seasonal variability. It serves as a benchmark for machine learning models in short- and long-term demand forecasting, microgrid management, economic dispatch, and techno-economic sizing of battery energy storage systems (BESS).

Files

Steps to reproduce

To reproduce and work with the processed dataset (processed_data), users can directly utilize the consolidated Excel files organized into annual subdirectories following the naming convention YYYY_MM_Processed_ISSO.xlsx. As raw data extraction requires private login credentials for the web portal (Datalog ISSO), the automated processing pipeline transforms raw measurements into the standardized processed format through the following sequence: (1) reading raw export files, stripping initial metadata header lines, and standardizing/translating all column headers from Portuguese to English using the mapped parameter correlation (e.g., converting "Inicio" and "Fim" to "UTC-3 Start" and "UTC-3 End", "Frequencia" to "Frequency", etc.); (2) normalizing sampling timestamps and creating UTC-synchronized time columns ("UTC Start" and "UTC End") repositioned at the table's start alongside local time (UTC-3); (3) converting direct and reverse active power parameters from Watts (W) to kilowatts (kW); (4) calculating cumulative active energy in kilowatt-hours (kWh) for both direct flow (energy consumed from the utility grid) and reverse flow (surplus photovoltaic energy exported to the grid) based on the 15-minute sampling interval; and (5) running a four-step automated validation protocol that evaluates chronological continuity (flagging time gaps where the time interval are greater than 15 min), filters operational voltage violations outside ANEEL PRODIST Module 8 thresholds (117–133 V), verifies total three-phase active and apparent power balance consistency, and performs Ohm's law cross-checks (flagging relative error > 1.0%), saving all output logs in the validation_logs folder. Researchers can load these cleaned and validated .xlsx files directly into data analysis environments (such as Python/Pandas or R) to reproduce machine learning model training, short- and long-term demand forecasting, microgrid management simulations, and BESS sizing studies.

Institutions

  • Instituto Federal de Educacao Ciencia e Tecnologia do Espirito Santo
    ES, Vitoria

Categories

Measurement of Electrical Quantity

Funders

Licence