Rethinking data requirements for the reliability assessment of medium voltage cables

The main aging and failure drivers of medium voltage cables are known. However, the distribution grid's enormous number of cables makes it difficult to determine the status of each cable section individually. Machine learning approaches to predict the reliability of medium voltage cables are promising, but often use non-standard features. As a result, it remains difficult to quantify the influence of each failure driver and evaluate correlations. Therefore, this work redefines the data requirements for data driven approaches of reliability assessment for medium voltage cables, by providing an overview of features to represent aging drivers. Furthermore, main data sources for Denmark are identified and merged to assess issues in data collection, availability, and combination. Finally, data management, and feature selection tasks are discussed to accurately employ the defined data requirements in future condition monitoring applications. 

Consequently, this paper aims to contribute to the reliability assessment of MV cables, by:

1. Qualitatively define the features to model failure drivers, based on existing standards and literature
2. List main data sources of the defined features for Denmark
3. Evaluates challenges in terms of data collection, availability, quality, and combination
4. Provide an outlook on required feature selection processes and usages within ML models