Data-based Condition Prediction of Medium-Voltage Substations for Optimised Asset Management Strategies Supported by Artificial Intelligence Methods

Inhalt:
The use of artificial intelligence for optimisation of maintenance and renewal strategies concerning medium-voltage substations in the distribution grids is investigated. An important aspect here is the expected condition development of the medium-voltage substations and its components. The concentration in this work lies in the condition prediction for the medium-voltage switchgear. To test the data-based condition prediction, a standardised data model for master data and condition data is developed based on datasets from two distribution system operators. On this basis, a weighted condition assessment is applied and a normalised condition index for the medium-voltage switchgear is determined. As part of the analysis, the quantity structure and the data quality are analysed to evaluate the possible use of machine learning for the derivation of further switchgear condition development and to identify the most significant parameters. The switchgear age is identified as a significant factor correlating with the condition development as well as the number of days since last maintenance. The first prediction results show already a promising performance. In the further course of this research project, data from more distribution system operators will be included for covering a broader input database and for reaching a higher prediction accuracy.