Application of Optimal Multiplier Method in Weighted Least-squares State Estimation Part Ii: Simulation

Standard algorithms for state estimation may be viewed as quasi-Newton's methods applied to the rst order optimality conditions of a least squares minimization problem. Previous work in the literature has documented the (somewhat surprising) fact that when a full Newton's method is applied to the same formulation, convergence properties are far worse than the quasi-Newton's method, until the iterates reach an EXTREMELY small neighborhood of the solution. Motivated by these results, and by availability of e cient algorithms to compute higher order derivatives necessary in an exact Newton formulation, the companion paper [2] proposes several Newton's method variants to improve state estimator convergence. In this paper Benchmarks for the IEEE 118 and 300 bus test systems are provided, with comparisons against classical normal equations, Hatchel's method, and QR algorithms. In these benchmark examples, the new algorithms developed show more reliable convergence for ill-conditioned cases, while making minimal sacri ces in computational e ciency for well-conditioned cases.