Moritz Cramer

Bad data detection and handling in distribution grid state estimation using artificial neural networks

This research project addresses the problem of erroneous measurements for distribution grid state estimation (DGSE) by using bad data detection. A method based on artificial neural networks is developed and tested in combination with DGSE. The necessary steps in order to create the neural networks are presented and training parameters are investigated. The method replaces identified measurement errors with new estimates. The developed method is validated by simulation on the basis of an electric distribution grid. It is shown that the bad data correction method detects and correctly identifies single and multiple erroneous measurement values. Furthermore, the ability of the developed method to handle different types of measurement errors and the impact on the quality of the state estimation result are examined. The method improves the quality of the DGSE result and reduces the state estimations probability to diverge.

Optimal use of decentralized methods for volt/var control in distribution networks

This paper presents results of a heuristic optimization method for parameterization of decentral reactive power control, which can be used to improve static voltage quality in distribution grids. With an optimized Q(V)-curve it is possible to significantly reduce the additional reactive energy and accomplish lower losses in the grid while keeping the voltage within the specified limits. In this paper results for Q(V)-control and optimized cosφ(P)-curves with optional voltage regulated distribution transformers are presented. It is shown that the gradient of a Q(V)-curve should be as steep as possible. The optimization method also shows positive impact of bigger reactive power capability of each DG unit. Individual Q(V)-parameters for a group of distributed generators can significantly improve the efficiency of Q(V). The comparison with results using Optimal Power Flow shows a significant gap between optimal decentralized methods and coordinated optimization. Using coordinated optimization it is possible to reduce the yearly reactive energy of all DG units by over 95%.