Neural network modeling of STATCOM using the GAMMA and the RBF identifiers

Analysis of power systems is carried out in frequency-domain (e.g. load flow studies). Currently FACTS controllers are included in the analysis, whereas power electronic-based devices are modelled in the time-domain using state-space equations (e.g. STATCOM). Since these models cannot be directly used for power system analysis, it is necessary to seek suitable alternatives that are applicable to the frequency-domain analysis. Average model of STATCOM is a time-domain representation in which high frequency switching ripples are vanished. This paper extends existing average model, and presents an average-neural model of STATCOM to bridge the time and frequency domains analysis. Thus, development of two identifier neural networks is studied, namely Gamma and radial basis function (RBF). Then, these two models are compared with their exact solutions in terms of the least average error. Selected average-neural model is then used as an application in the 14-bus IEEE power system to verify the usefulness of the suggested model for power system studies.