Development of a Risk-Based Health Index for High-Voltage Assets to Enhance Asset Management

The energy transition requires a transformation of the electrical transmission grid. Grid expansion and more targeted utilisation of the grid are the main countermeasure to ensure system reliability. However, these measures are accompanied by limited resources and increasing demands for controlling and maintaining the grid. Consequently, there is a growing need to gain a deeper understanding of the condition of each asset and its impact on system operation. In this context, asset management (AM) offers a structured framework to optimise asset utilisation and replacement in a value-based manner. However, a survey and literature review conducted as part of this thesis show that asset management (AM) practices among electric grid operators (EGOs) are diverse with varying levels of maturity and different advantages and limitations. This thesis addresses these challenges by analysing and comparing various asset assessment methods, using circuit breakers and power transformers as two representative practical examples. The objective is to provide a comparative foundation that supports the AM based on available data and specific asset characteristics. In this thesis, a risk-based health index (HI) customisable to the needs and constraints of typical electric grid operators (EGOs) is developed. The health index (HI) is based on a flexible four-level assessment structure, which allows for triggering and prioritising maintenance measures using asset risk. As a result, maintenance measures can be objectively compared based on their impact and effort, allowing for quantitative and value-based decisions. At its core, the proposed HI uses Bayesian probability to model asset risks, combining prior probabilities with available evidence. The Bayesian model allows for the consideration of various kinds of evidence and their combination. The level of detail can be expanded to quantify the influence of a single measured value. To improve the accuracy of the failure probability estimation and to make better use of typical EGO data, three methodological extensions are introduced: • Determination of dataset-based prior probabilities using Weibull distributions and machine learning regression. • Classification of measurement data into condition states using machine learning classifiers, enabling the derivation of conditional probabilities for Bayesian updating. • Estimation of a technical age to bundle ageing processes for a facilitated representation of ageing in the probability assessment. The resulting HI offers an adaptable assessment framework that enables more accurate and objective maintenance planning under uncertainty and limited data availability. It supports consistent risk-based decision-making across asset types, facilitates digitalisation, and helps EGO increase the targeted and efficient utilisation of their electrical grid.