Detecting, locating and classifying earth faults
A large part of the Norwegian low voltage distribution network is of the type 230 V IT-grid, in contrast to most other countries in the world, which mostly have TN-grids [Sikker jordfeildeteksjon; Helge Seljeseth og Eilif Hugo Hansen]. An advantage with the IT-grid is that the current is small during earth faults, and therefore these types of grids can operate with persisting earth faults. This means that in locations where the electricity supply is crucial (for example hospitals) an earth fault will not be as critical.
Today, the DSOs often detect and locate earth faults in their LV distribution network when they receive customer inquiries about it. However, earth faults can be hard to discover for customers as well.
The DSOs are required (According to FEF2006 from DSB) to monitor earth faults in selected substations in their network. Earth faults are detected by analysing the voltage in the substation that is measured by multi-instruments and then using a filter. The filter analyses all variations in phase voltage and classify the ones which exceed a standard value given by REN as earth faults. Transfer of data is made through the smart meter infrastructure.
The deployment of smart meters to all customers in Norway is a unique opportunity to achieve a better overview of earth faults. Normally, the sum of currents on the three phases in the LV distribution network is equal to zero. However, during an earth fault it is not equal to zero. The smart meters can be used to calculate this sum, and together with the analysis of voltage in substations, earth faults can be located by "asking" the smart meters under a substation where an earth fault has been detected what their respective sum of currents is. An analysis of AMS data using machine learning can in addition give the
type of earth fault and classify which appliance is causing the earth fault.