Data Driven Detection of Malfunctioning Devices in Power Distribution Systems Validation (DeMaDsVal)

Aselectricity grid operators encounter new challenges in grid operation due to profound changes in the electric energy system, such as decentralization of generation, also new methods to cope with these challenges are sought after. Therefore, an investigation of a concept for remote de tection of malfunctioning grid-supporting devices is under development within the project. The operation of future electricity grids depends on the behavior of these devices and their sup port functions such as reactive power dispatch, used for example for voltage control. Using operational data of medium voltage transformers at first, as well as topological data and smart meter data at the low voltage level, the functionality developed is to enable better surveillance of grid-connected devices. This is to be achieved by combining machine learning algorithms for anomaly detection, classification, and load disaggregation. These are applied to the trans former data as well as to the device data to identify and classify unwanted behaviour. The aim is that the framework should be a future tool for grid operators and for cooperation with them to help them implement a central novel surveillance of low voltage grids regarding the connected devices. This framework will also be tested with some selected use cases in order to prove its usability. The data used will both be generated synthetic data from grid simulations as well as recorded data that can be gained in laboratory setups. The data collected in laboratory scenar ios can then on the one hand be used to further enhance the quality of the synthesized data by comparing and filtering out possible influence factors that might have been neglected in the simulations.

On the other hand, the data can be used as a validation set to validate the performance of the used machine learning methods. These are trained and tested on the synthetic data, making such avalidation set very valuable to assess the robustness of the approach and also be able to further improve the same. Multiple scenarios and setups were implemented to capture various use cases under different circumstances. The outcomes of the work are therefore the collection of such a validation set of operational data of grid participants and substations in scenarios that involve misconfigurations of grid connected devices such as inverters, battery energy storages or controllable loads. This dataset as a main outcome will then be used to robustify and further develop the monitoring approach.