The student Ibai Ramirez Garcia obtained an GOOD with mention INTERNATIONAL DOCTORATE

The student Ibai Ramirez Garcia obtained an GOOD with mention INTERNATIONAL DOCTORATE

---

• Thesis title: Transformer Insulation Heath Monitoring Framework for Improved Lifetime Management through Physics-Informed Machine Learning Methods

Court:

• Presidency: Juan Carlos Ramos González (TECNUN)
• Vocal: Kateryna Morozovska (Imperial College London)
• Secretary: Mikel Mendicute Errasti (Mondragon Unibertsitatea) 

Abstract:

The increasing use of renewable energy sources (RESs) with highly dynamic power generation, has created previously unforeseen scenarios, which affect the lifetime estimation of the power grid assets. The transformer is an essential asset for the correct and reliable operation of the power grid, and its most relevant point of failure arises from the thermal degradation of its insulation. Transformer insulation degradation is most critical when the temperature of the insulation is highest, which is known as the winding hottest-spot temperature (HST). Accordingly, for the correct and reliable operation of transformers, it is crucial to monitor the transformer HST. However, as the HST moves spatially over time, direct HST measurement is complex.

In this context, the main objectives of this thesis are focused on the use of advanced Machine Learning (ML) models, along with physics-based models, for improved transformer lifetime management. The main development applications focus on (i) intelligent feature selection, (ii) improved transformer thermal modeling, and (iii) uncertainty estimation of measurements, using the developed physics models:

1. For the reliable and efficient operation of RES it is crucial to monitor the health of its constituent components and feature selection is a crucial step for building robust and accurate health monitoring approaches. In this context, this research presents a probabilistic feature selection approach, which probabilistically weights and selects features through a heuristic and iterative process for an improved asset lifetime estimation. The approach has been applied to transformer lifetime estimation, but it can be generally applied to assist in the lifetime estimation of other components operated in RESs.
2. HST measurement is complex, costly, and often estimated from indirect measurements. Existing HST models focus on space-agnostic thermal models, providing worst-case HST estimates. This research introduces a spatio-temporal model for transformer winding temperature and ageing estimation, which leverages physics-based partial differential equations (PDEs) with data-driven Neural Networks (NN) in a Physics Informed Neural Networks (PINNs) configuration to improve prediction accuracy and acquire spatio-temporal resolution. Furthermore, the spatio-temporal transformer ageing model is inferred, which supports transformer health management decision-making.

(iii) Bayesian Physics-Informed Neural Networks enhance prognostics by integrating physical laws with data while jointly modeling epistemic and aleatoric uncertainty. This enables more accurate predictions, well calibrated uncertainty estimates, and supports risk aware, uncertainty informed decision-making in transformer health management.