Data-Driven Dynamic State Estimation of Synchronous Generators via Sparsity Promoting Unscented Kalman Filtering

This paper introduces an innovative model-free and data-driven approach to dynamic state estimation in nonlinear generators. The proposed framework is structured around a two-phase process comprising data-driven model identification and state estimation. In the initial phase of model identification, conducted within a short time frame (a few seconds), we gather state feedback to elucidate the dynamics of nonlinear generators within the power grid. This is accomplished through a modified sparse regression algorithm, distinguishing it from conventional methods that rely on extensive libraries of linear and nonlinear functions. Our approach adopts an iterative strategy, commencing with a compact library and incrementally incorporating higher-order nonlinear functions based on coefficient density. Having identified the system’s dynamics, the subsequent dynamic state-estimation (DSE) phase identifies the states of the generator from limited and noisy measurements using a data-driven unscented Kalman filter approach. The effectiveness of the proposed sparse regression unscented Kalman filter is demonstrated through simulation results, validating its effectiveness for a model-free state-estimation of synchronous generators in power grid. The computational and memory efficiency of the proposed method is demonstrated through a comparison with conventional sparse regression.