A Subspace-Guided Constrained Optimization Framework for M-Class Synchrophasor Estimation Under Nonstationary Conditions

In recent years, the integration of renewable energy sources and the widespread use of nonlinear loads have increased dynamic uncertainties in modern power systems, making real-time and synchronized monitoring essential. Accurate M-class synchrophasor estimation under these nonstationary and spectrally uncertain conditions remains a challenging problem due to dynamic variations, harmonics/interharmonics, out-of-band interference, and measurement noise. This study proposes a suitably constrained optimization-based framework for M-class synchrophasor estimation, in which a hybrid structure integrating an ESPRIT-based subspace method with the Adaptive Fitness Distance Balance Artificial Rabbit Optimization (ES-AFDB-ARO) algorithm is employed. In this framework, the optimization stage is guided by spectral information obtained via the subspace stage to narrow the search space and improve convergence stability. Performance is evaluated under IEEE C37.118 steady-state and dynamic conditions via Monte Carlo simulations, showing that total vector error, frequency error, and rate-of-change-of-frequency error values remain within standard limits. Comparative analyses at 60 dB and 40 dB SNR demonstrate that the ES-AFDB-ARO method exhibits improved and more stable performance than the widely used interpolated discrete Fourier transform, Taylor weighted least squares and Taylor–Kalman filter methods. The results show that the proposed framework offers a reliable solution for synchrophasor estimation under dynamic operating conditions.