Multi-Device Data Analysis for Fault Localization in Electrical Distribution Grids
Abstract
The work presented in this dissertation represents work which addresses some of the main challenges of fault localization methods in electrical distribution grids. The methods developed largely assume access to sophisticated data sources that may not be available and that any data sets recorded by devices are synchronized. These issues have created a barrier to the adoption of many solutions by industry. The goal of the research presented in this dissertation is to address these challenges through the development of three elements. These elements are a synchronization protocol, a fault localization technique, and a sensor placement algorithm.
The synchronization protocol addresses the dependency on synchronized data by allowing the devices themselves to synchronize their data. This is accomplished by establishing relationships between temporal events, transient signals and the devices recording these events and signals. The protocol establishes a relationship between transient signals emitted from standard equipment in electrical grids and the fault event which is then leveraged to synchronize the recorded fault data. The method has shown very promising results in synchronizing data from multiple devices.
The fault localization technique determines the location of a fault in a complex distribution grid. This is achieved by establishing spatial-temporal relationships between transient signals and distances in graphical representations of the distribution grid. This method provides precise fault locations in challenging electrical distribution grid structures. It also allows areas of the distribution grid to be classified based on how well they are monitored. This classification component predicts how well a distribution grid is monitored by the devices used.
The sensor placement algorithm leverages the classification component established in the fault localization technique to optimize the placement of the fault localization devices in the distribution grid. Here, an efficient algorithm is created for determining the best location for a given number of fault localization devices.
The combination of the synchronization protocol, a fault localization technique and a sensor placement algorithm and their minimal data requirements constitutes a uniquely complete solution that may set it apart from existing work and make it more attractive to the industry for adoption in electrical distribution grids.