Avoiding Medium Voltage Cable Joint Failure: Development of a Real-Time Prognostic Tool

Cable failures are disruptive, costly to repair and have a serious impact on customer confidence. Thus, developing a reliable on-line prognostic tool is of a great interest. An experimental setup has been installed within a distribution substation in Cyprus for a period of 5 years. Temperature condition monitoring units were installed on selected underground cable joints. Real-time current loading data, weather conditions as well as surface temperature by the cable joints were used for the development of the thermal prognostic models which predict the likely temperature along the cable joint thirty minutes into the future. This paper introduces a series of novel thermal condition monitoring machine learning techniques that can be used to effectively monitor the performance of cable joints.