Towards a performance index for distribution transformer prioritization and management

• Proposes a data-driven performance index for distribution transformer prioritization. • Uses only commonly available utility data, avoiding advanced monitoring systems. • Captures utilization patterns beyond peak load and insulation aging metrics. • Combines feature engineering with PCA and autoencoder-based dimensionality reduction. • Validated on large-scale real data from Brazilian utility transformer fleets. Differentiating the operational condition of distribution transformers is essential for their effective management. Traditional approaches based on overloading or equivalent aging capture only limited aspects of operation and offer restricted support for management decisions. This paper proposes a performance index that prioritizes distribution transformers by quantifying key aspects of equipment operation through comprehensive feature engineering applied to data commonly available to utilities. Using information such as rated capacity, active power demand, monthly energy, and consumer connections, 13 features were derived and incorporated into a two-step learning framework based on dimensionality reduction that produces a single valued and interpretable performance index. Results show that the index effectively separates units according to their utilization patterns and provides a practical decision support tool for targeted management interventions.