A new technique for unconstrained and constrained linear LAV parameter estimation

This paper presents a new technique for least absolute value (LAV) linear parameter estimation. The technique can be applied to unconstrained as well as equality- and/or inequality-constrained problems. The estimation problem is stated first. The least squares (LS) solution of the estimation problem is then presented and compared to the LAV solution. The new LAV technique is offered and compared to the standard LAV algorithm which uses linear programming (LP). Several examples which compare results obtained using the LP and this new technique are presented. The examples provide the basis for the conclusions presented in the final section of the paper. The preliminary application of the new algorithm to the power system state estimation problem is also given.