Robust parameter estimation for fuel cost models in economic operation of electric power systems

This paper is focused on optimal parameter estimation of fuel cost models intended for economic scheduling of electric power systems. The common approaches of least squares and conventional weighted least squares may not be appropriate for situations involving non-normally distributed measurement errors. Techniques of robust parameter estimation, in particular the iteratively reweighted least-squares estimator, offer an attractive option in these situations. The paper discusses the theoretical aspects of this powerful technique and presents computational experience with a number of popular choices of weighting functions. A comparative evaluation of the parameter estimates obtained is carried out in terms of both accuracy of the estimates and their effects on economic scheduling strategies.