Automatic Voltage Regulator Model Validation Based on Mean-Variance Mapping Optimization and Field Tests

Abstract

Power system planning and operation are meaningfully based on a number of analyses which entail steady-state and dynamic simulations. In this regard, modelling the power system with enough detail is a basic requirement mainly for those applications that are based on an accurate prediction of the system dynamic response, such as the design of protective strategies and control schemes. Most models correspond only to a mathematical representation, whose parameters need to be firstly adjusted or identified based on a rational process of model validation which frequently employs experimental data. This paper proposes a parameter estimation method for accomplishing the model validation of power systems through an iterative software-in-the-loop (SIL) simulation, implemented via mean-variance mapping optimization (MVMO) in DIgSILENT PowerFactory, which allows comparing the simulation results with records obtained from field tests. The proposed method is then used to perform the model validation of the automatic voltage regulator (AVR) of Coca Coco Sinclair, Ecuador’s largest hydroelectric power plant. The obtained results are finally compared with two other similar approaches: i) the “Model Parameter Identification” object of PowerFactory, and ii) the “Parameter Estimation” toolbox of MatlabSimulink. Comparisons show the benefits of the proposal to overcome limitations of the other two methods regarding accuracy, constraints and SIL simulation capabilities.