The effect of normalisation on the performance of a stochastic power system load model

Previous work has shown that an ARMAX based time series dynamic load model is robust in predicting load dynamics for transient events at any seasonal time and location. This paper extends this work by exploring the effect of the normalisation of measured transient data on the performance of ARMAX load models, for a set of transient events at a 66 KV substation in the Victorian power system. The simulations compare the output of an ARMAX load model developed using normalised values of measured transient data, with that of an ARMAX model built with absolute measured transient data. The results show that the model built with normalised data exhibits much better load response prediction performance when compared against measured data from transient events other than that from which the model was constructed. The paper then explores the appropriate base quantity to be used for the improved performance of the normalised model and proposes a strategy to incorporate the normalised model into PSS/E. Finally, the performance of the normalised model for one transient event is compared with that of a composite load model taken from within the PSS/E analysis package.