Power system state estimation using l(1) norms and interior point methods

While considerable time has elapsed since Weighted Least Absolute Value (WLAV) state estimation was first proposed as an alternative to Weighted Least Squares (WLS) state estimation for the real-time control of power systems, its viability for on-line application still remains to be fully established. This thesis describes recent contributions towards establishing such a viability by using interior point methods for WLAV state estimation. The use of interior point methods not only makes WLAV state estimation computationally faster, but also makes it simpler to implement by translating the problem into the solution of a sequence of least squares problems. These least squares problems are identical in structure to those in conventional WLS state estimation. This enables the use of well known methods that exploit the sparsity pattern of the power system state estimation problem. As a consequence, it becomes easier to integrate a WLAV option in the state estimation function in an energy management system. The dual formulation of the WLAV estimator is shown to have definite advantages over the primal formulation for application of interior point methods.; Techniques to exploit the nature of WLAV estimators for obtaining inexpensive approximations to the true WLAV solutions are also explored. These approximate WLAV solutions can provide faster tracking of the system state and be useful to system dispatchers if they are displayed along with more accurate estimates which are updated on a slower time scale.; The problem of topology error detection has received little attention in the state estimation literature and still remains to be resolved satisfactorily. A new technique for determining the topology of electric power networks based on WLAV estimators is introduced in this thesis. This technique is based on the parameterization of all unknown admittances by 0-1 integer variables that are subsequently estimated by the WLAV estimator. The solution of the underlying combinatorial problem by means of both branch and bound methods and genetic algorithms is also discussed.; The use of WLAV estimators for external system estimation is explored. This involves the use of an exact penalty function. Comparisons of this technique are made with the fuzzy linear programming approach. Comparisons are also made with constrained WLS estimators based on interior point methods.; Equivalence between different formulations of the WLAV problem is established using the concept of minimization models.