Application of artificial intelligence for the thermal rating of a high voltage cable system

This paper briefly summarises the investigations carried out by Wiener Netze on the application of artificial intelligence for the thermal rating of high voltage cables in urban areas. As part of this project, various stationary and dynamic load tests were carried out on a 400 kV cable installation over a period of three years. In this time more than 70 million data points of various parameters such as currents, temperatures, soil moistures and climatic quantities were recorded. Based on this data, the thermal dynamics of the cable system were analysed using descriptive statistics. Established calculation methods were verified and, among other things, machine learning models were also trained and validated. It could be shown that machine learning models can recognise and depict complex processes that influence the thermal behaviour of the cable system using data alone. Furthermore, data-driven models were created that can not only map the cable temperature but can also be used for cable temperature forecasts for different time horizons within intraday forecasting with over 99 % accuracy. The lack of robustness of the black box models could be countered by appropriate validation methods for the cable arrangement under investigation.