Handling dynamic uncertainties for electric power system modeling and control

Abstract

This paper demonstrates an approach to treat partial dynamic uncertainties for purposes of electric power system monitoring and control. Stochastic modeling from real time measurements, and system parameter estimation techniques are used to identify the unknown dynamics. Cases of dynamic uncertainties, often encountered in electric power system studies, and discussed in this paper involve: [i] Electric load demand forecast, and/or system frequency prediction for purposes of system planning and automatic generation control. [ii] Dynamic characterization of electric loads or generators for different stability studies. [iii] Equivalencing of unknown parts of partially specified networks for handling large scale system complexities, computational cost and storage limitations. For a given sequence of load demands, Case (i), or frequency, a stochastic one-step-ahead predictive model is postulated and a set of undetermined parameters is estimated. In Case (ii), dynamical model for transients as seen at a particular load bus is formulated from current and voltage data gathered. Two separate feed-forward and feedback-loops generating current and voltage processes are then identified, conveying the interdependence of the two processes and orthogonality of disturbances. In Case (iii), dynamic equivalencing of totally or partially unknown parts of a large power system is considered. The static performance of the unspecified part is represented through generalised Thevenin's theory, or computed from tie line measurements. The dynamic impact of the unknown areas on the parts of concern is analyzed as a stochastic stationary process excited by unknown disturbances and observable inputs. Some case studies are referred to.