Dynamic State Estimation for Power Systems with Uncertain Inputs

Traditional dynamic state estimation (DSE) techniques focus on the filtering of noise and gross errors in output variables, which are commonly assumed to come from measurements. In power systems, however, input variables may have substantial uncertainty as well, as many of them are also measured or telecommunicated signals. This article discusses the impact of uncertainty in different types of inputs and proposes an adaptive iterated cubature Kalman filter with an uncertain input (AICKF-UcI) approach to systematically handle this problem. In the prediction stage, the method utilizes the Cubature transform to address system nonlinearity. Then, it converts the correction stage into a problem akin to weighted least-squares (WLS) regression, enabling iterative joint estimation of the state and uncertain input. Additionally, this method incorporates an adaptive algorithm for real-time estimation of noise distribution. Furthermore, the method is devised to have the unique capability of detecting and suppressing gross errors in input variables, and differentiating them from those in output variables. The advantages and versatility of the proposed method are validated through DSE on a permanent magnet synchronous generator (PMSG)-based wind turbine in a distribution system and a synchronous generator (SG) in a transmission system.