Additive Weibull Survival Analysis of Distribution Transformers in the Presence of Censoring and Truncation

The growing integration of renewable energy sources into distribution networks leads to changes in grid operations, especially with respect to load profiles, which may influence the aging and failure rates of medium voltage (MV) distribution transformers. This paper presents a survival analysis of MV distribution transformers using the additive Weibull distribution, which effectively models the entire asset life cycle, including infant mortality, random failures, and wearout phases, which cannot be captured by a standard Weibull model. The study addresses the challenges posed by historical data, including right-censoring, interval-censoring, and lefttruncation, which are common in real-world asset datasets. Using both synthetic and actual data from a Dutch distribution system operator, the impact of these censoring mechanisms on parameter estimation is analyzed and corrected. The Turnbull estimator is used to validate the shape of the survival function. Additionally, stratification by voltage rating highlights the bias introduced by uncorrected truncation. The results demonstrate that correcting for left-truncation significantly improves the fit of the model and reduces survival overestimation. This work contributes to the development of data-driven reliability models and highlights the importance of properly addressing data limitations in power component reliability modeling to avoid overestimating component survival time to support effective asset management.

Recommended citation: N. Brekelmans, Y. Zhang, P. H. Nguyen, A. van der Molen and P. van der Wielen, "Additive Weibull Survival Analysis of Distribution Transformers in the Presence of Censoring and Truncation," 2025 IEEE International Conference on Environment and Electrical Engineering and 2025 IEEE Industrial and Commercial Power Systems Europe (EEEIC / I&CPS Europe), Chania, Crete, Greece, 2025, pp. 1-6, doi: 10.1109/EEEIC/ICPSEurope64998.2025.11169264.
Download Paper