Reliability models of distribution networks for shorter time scale have never been established in the past due to lack of data. Over the years DE has collected extensive data from the distribution grid operators in Denmark that records the steady state and fault events in the distribution grids. Other data sources are also available, such as weather observations, GIS, and digging permits, which may influence the likelihood of component failure. This provides an opportunity to build a reliability model of the network and explore the correlations between data.
The model can be used to predict the probability of future events, and hence plan the grid maintenance schedule optimally. The work will further explore the cost and benefit of different options for the grid operators, such as where to add sensors or automated systems to isolate and recover from faults. The model will be created using deep reinforcement and unsupervised learning techniques from the databases.
Research Directors:
