Real time instability prediction through adaptive time series coefficients

A method for predicting, in real-time, the outcome of evolving transient behavior in power systems through adaptive time series coefficients is presented. After the start of a power swing, an auto-regressive model is fitted to the generator power angle in order to predict the outcome of that swing for future intervals. In this research, to estimate the parameters of the model, data are sampled at a rate of 100 times per second. These measurements may be obtained from the synchronized phasor measurement units (PMUS). Data of an observation window from one to two cycles of sixty Hertz are used to estimate the parameters of the model. In order to have an adaptive scheme, new parameters of the model are estimated once the prediction step is completed. The process is activated only in case the generator angle measurement between one step and the next differs by more than a threshold value.