Power grid modelling from wind turbine perspective using principal component analysis

In this study, we derive an eigenvector-based multivariate model of a power grid from the wind farm's standpoint using dynamic principal component analysis (DPCA). The main advantages of our model over previously developed models are being more realistic and having low complexity. We show that the behaviour of the power grid from the turbines perspective can be represented with the cumulative percent value larger than 95% by only 4 out of 9 registered variables, namely 3 phase voltage and current, frequency! active and reactive power. We further show that using the separation of signal and noise spaces, the dynamics of the power grid can be captured by an optimal time lag shift of two samples. The model is finally validated on a new dataset resulting in modelling error residual less than 5%.