Application of Bayesian method for electrical power system transient stability assessment

In this article we present a short-term (short-transient) dynamic stability assessment methodology for generators of a power system at characteristic/relevant operation modes. The suggested methodology combines a classical simulation methodology and Bayesian approach, the latter being widely applied in a variety of areas nowadays. The proposed methodology reduced the number of calculations and the respective cut in computational time, enabled obtaining quite precise results in case of the lack of real measurements. The idea was to replace the major portion of calculations with a mathematical model of the power system by estimation procedure. As for the specification of its features, the methodology takes into account characteristic operation modes of the system in annual cycle and uses dependency functions of the generators relating dynamic stability reference characteristics with the estimated characteristics. The estimated dynamic stability characteristics of the same generators were found with the Bayesian-approach-based estimation model which was trained by data sequence of true points from the reference characteristics. The performance of the methodology was examined by testing the dynamic stability of the generators of the power system and statistically analysing the test results. In summary, it can be reasonably argued that the best estimates were produced when the estimation model was trained with one dynamic stability reference characteristic and three true points of the characteristic for the considered generator. The verification of the suggested methodology for the aforementioned test system revealed that it reduces the computational time by several times, in comparison to "classical" methodology (based only on mathematical modelling of the power system). (C) 2012 Elsevier Ltd. All rights reserved.