SALTLoad: Distribution Substation Electrical Load Dataset

Published: 30 July 2026| Version 2 | DOI: 10.17632/vgmxfs4yk9.2
Contributors:
,
,
,

Description

This repository contains SALTLoad, a high-resolution multi-feeder electrical load dataset collected from a distribution substation operated by the Saltha Palli Bidyut Samity in Faridpur, Bangladesh. The dataset consists of synchronized electrical measurements recorded at 5-minute intervals from six distribution feeders and is intended to support research in electrical load forecasting, anomaly detection, energy analytics, and data-driven power system applications. For each feeder, the dataset includes three-phase voltage, three-phase current, active energy, reactive energy (import and export), apparent energy, and derived power factor measurements. The electrical measurements are integrated with hourly meteorological observations, including temperature, humidity, precipitation, wind speed, solar radiation, atmospheric pressure, and cloud cover, together with calendar-based annotations identifying weekdays, weekends, public holidays, and major festivals. Five feeders (Feeders 1, 2, 4, 5, and 6) span 1 August 2021 to 31 May 2026 (approximately five years), while Feeder 3 covers 1 August 2021 to 30 April 2025, after which it was transferred to an externally managed system and no further measurements were available. Each complete-span feeder contains up to 508,033 synchronized 5-minute observations, whereas Feeder 3 contains approximately 393,985 observations. The repository includes the original raw data together with successive preprocessing versions documenting data cleaning, timeline reconstruction, weather integration, holiday annotation, multi-stage missing data recovery, and the final curated dataset, providing a transparent and reproducible data processing pipeline for future research.

Files

Steps to reproduce

CSV files are the RAW dataset files. The README file contains the concise information about the dataset and how to use it.

Categories

Electrical Engineering, Machine Learning, Energy Consumption, Time Series Forecasting, Energy Forecasting, Deep Learning

Licence