Fig 3 - uploaded by Konrad Sundsgaard
Content may be subject to copyright.
Source publication
The digitalization of the power distribution grid has surged over the past decade. This transformation has given rise to a host of new data-driven applications focused on condition monitoring and predictive maintenance. However, from the perspective of the distribution system operator, there remains uncertainty about what and how digital maintenanc...
Context in source publication
Context 1
... useful life prediction, and others. To sharpen the taxonomies used, this study narrows down the application parent categories to the three classes: Condition monitoring, Ranking, and Prognosis. Early fault detection is thereby listed as a subset of condition monitoring. Remaining useful life prediction is seen as a special type of prognosis, see Fig. 3. Hence, the authors acknowledge that other divisions or combinations of the above might be feasible as well. The result from condition monitoring will, for example, often be included as an input along with other condition parameters in ranking or forecasting. Furthermore, in principle, a ranking can also be based on a remaining-life ...
Citations
... These networks usually comprise older components, are less accessible than their high-voltage counterparts, and significantly contribute to grid outages. Previous research has predominantly concentrated on high-voltage assets and sensor-based solutions, underscoring the need for a targeted exploration of distributed, underground assets like MV cables [3]. In Denmark, the two primary cable technologies-PILC (paper-insulated leadcovered) and XLPE (cross-linked polyethylene)-exhibit significantly different failure profiles. ...
... Furthermore, (3) shows that L cat equals the sum of the crossentropy losses computed for each of the d cat categorical variables. Consequently, it measures the discrepancy between the true categorical values x cat,k and their predicted probabilitieŝ x cat,k . ...
Electricity distribution cable networks suffer from incomplete and unbalanced data, hindering the effectiveness of machine learning models for predictive maintenance and reliability evaluation. Features such as the installation date of the cables are frequently missing. To address data scarcity, this study investigates the application of Variational Autoencoders (VAEs) for data enrichment, synthetic data generation, imbalanced data handling, and outlier detection. Based on a proof-of-concept case study for Denmark, targeting the imputation of missing age information in cable network asset registers, the analysis underlines the potential of generative models to support data-driven maintenance. However, the study also highlights several areas for improvement, including enhanced feature importance analysis, incorporating network characteristics and external features, and handling biases in missing data. Future initiatives should expand the application of VAEs by incorporating semi-supervised learning, advanced sampling techniques, and additional distribution grid elements, including low-voltage networks, into the analysis.