Longitudinal Multi-Source Monitoring Dataset for Power Transformer Health Index and RUL Prediction

The current dataset presents longitudinal monitoring data and engineered features for 166 oil immersed power transformers over a period of 10 years. The dataset is divided into two separate CSV files that serve to support predictive maintenance modeling, condition assessment and asset management analysis as follows: File 1: Basic Features. This file consists of historical field data at a component level that include multi-source diagnostic parameters. These consist of Dissolved Gas Analysis (DGA), oil quality parameters, paper insulation degradation parameters and operational temperature parameters, thus providing a versatile baseline of physical state representation. File 2: Engineered Feature Dataset. This file contains the baseline real world data alongwith engineered features created through this research. This includes health index derivative features, multi-variable thermal stress interaction features and filtration based historical maintenance variables. The dataset serves to train algorithms, test robustness in data sparse conditions and LCC optimization applications.