SOLUÇÃO DE SISTEMAS DE EQUAÇÕES DIFERENCIAS E ALGÉBRICAS: APLICAÇÃO EM SISTEMAS DE ENERGIA ELÉTRICA

Abstract

The present work investigates and compares the computational performance of numerical techniques, selected from specialized literature and applied to the solution of large-scale algebraic and differential equations. Among the considered techniques, emphasis was given to the method known as MEBDF (Modified Extended Backward Differentiation Formulae), because it presents properties that the conventional BDF (Backward Differentiation Formulae) methods do not have, and these properties may improve its computational performance in certain applications. The numerical methods considered in this research are available as computational numerical codes, known as solvers, and they are of public domain. The ones considered here are the MEBDFSD, MEBDFI, DASSL and RADAU. The computational tests considering these numerical codes are related to simulations of power system transient angular stability and long-term voltage stability in the time domain. The main objective was to check the efficiency of these numerical techniques under two aspects, namely, the computational efficiency and numerical accuracy. The computational aspect is related to the simulation CPU time, and accuracy is related to the obtained numerical results, since these methods use, in general, approximation techniques. A conventional stability program was used to validate the results. Computational analysis was performed using the following test systems: IEEE118 buses with 54 generators, IEEE145 buses with 50 generators, and an equivalent south-southeast Brazilian power system. The obtained results indicate that the MEBDFSD performance is better rather than the other methods considered here.