A comparison of algorithms for least absolute value state estimation electric power networks

We compare in detail the performance of three methods for solving a nonlinear programming formulation of the weighted least absolute value estimation problem; least squares is used as a benchmark. Two primal and one dual interior point technique are considered. The former are a primal-dual method and Mehrotra´s predictor-corrector method, both as implemented by Lustig, Marsten, and Shanno (1990). The latter is an affine scaling dual method as implemented by Alvarado and Singh. Testing on networks ranging from 6 to 828 buses with a variety of measurement sets shows that the predictor-corrector method is the most efficient and robust of the least absolute value techniques considered. The dual is a close competitor