Online prediction of power system trajectories from noisy data by penalized least-squares minimization

This paper advances a new method to predict trajectories of rotor angles and bus voltages in dynamic power systems, with data collected by phasor measurement units. The predictor incorporates real time data adaptively, leveraging the algebraic differential equations describing the system, while being able to accommodate model inaccuracies. Numerical tests on a one machine infinite bus system are carried out to illustrate the predictor properties. Then, further results are presented on a classical networked testbed of nine buses and three generators.