Data-Driven Approaches for Monitoring of Distribution Grids
Aachen / E.ON Energy Research Center, RWTH Aachen University (2017) [Book, Dissertation / PhD Thesis, Report]
Page(s): xii, 108 Seiten : Illustrationen, DiagrammeReport number: D-82
Electric power distribution systems are undergoing dramatic changes due to the ever-increasing power generation at the medium and low voltage grids and the large-scale grid integration of electric vehicles. This has led to the operation of distribution systems in a very different way compared to what they were originally designed for. Considering that little monitoring equipment is traditionally embedded in these systems, Distribution System Operators (DSOs) need cost-efficient monitoring methods to ensure the safe and efficient operation of their grids.In this dissertation, a data-driven approach to the monitoring of distribution systems is presented to allow DSOs to start monitoring their systems with minimal investments though with limited accuracy. DSOs can then incrementally improve the monitoring accuracy at a later stage if needed by adding further measurements. The proposed approach divides the distribution system into smaller sections and assigns a local estimator, based on Artificial Neural Networks (ANNs), to estimate node voltage magnitudes needed for the monitoring of each section. In addition, procedures are presented to determine the optimal selection order of the input candidates for the monitoring system to ensure the highest monitoring accuracy. The procedures extend the concepts of partial correlation and minimal redundancy maximum relevance (mRMR) to support problems with multiple target outputs. The proposed concepts are validated by being applied to test distribution systems.