Real-time implementation of high breakdown point estimators in electric power systems via system decomposition

This dissertation presents a new, highly robust algorithm for electric power system state estimation. A graph theory-based system decomposition scheme is coupled with a high breakdown point estimator to allow reliable identification of multiple interacting bad data even in cases of conforming errors. The algorithm is inherently resistant to bad measurements in positions of leverage, makes no a priori measurement error probability distribution assumptions, and is applicable in a real-time environment.

In addition to presenting a new state estimation algorithm, the weaknesses of two prominent state determination methods are explored. The comparative advantages of high breakdown point estimators are then summarized. New theorems quantifying the previously unexamined effect system sparsity has on the exact fit point of some members of this estimator family are presented. These results serve as the catalyst for the overall state estimation algorithm presented. Numerous practical implementation issues are addressed with efficient implementation techniques described at each step.