Out-of-step prediction logic for wide-area protection based on an autoregressive model

This work describes a new out-of-step prediction logic based upon the autoregressive model (AR model), which is a kind of time-series analysis. The oscillation after the clearance of a severe fault is modeled using the mth-order of the AR model, of which input data are the phase difference of the voltage between substations. The stability of the power system is distinguished using the sign of the real part of the characteristic roots that are given by solving the pulse transfer function of the AR model. That is, a negative sign leads the power system to be judged to fall into unstable status. In this case, the oscillation of the rotor angle of every generator is modeled using the mth-order of the AR model. Where both oscillations include the same components, the power system is predicted to fall into out-of-step status. The new logic is verified using the Y-method, which CRIEPI has developed as a dynamic analysis and which gives good results in the 10-generator western Japan 60 Hz model power system.