Research On Modal Extraction Of Electromechanical Oscillations In Power System Based On Subspace Dynamic Mode Decomposition

With the gradual expansion of the interconnection scale of multi-regional power grids,the structure of the power grid appears diversified and complicated,and the increase in the proportion of new energy output and the frequent random disturbances will bring challenges to the security and stability of the system,leading to frequent electromechanical oscillations that threaten the stable operation of the system.Therefore,no matter what stage the power system is in,it is necessary to evaluate and analyze the small signal stability of the system in time.Based on the steady-state measurement data under environmental incentives,this paper studies an improved method for extracting modal information of electromechanical oscillations.The specific research contents are as follows:First,the power spectrum density analysis results are used to distinguish the response types of the power system.The characteristics of the random response data with small fluctuations and the advantages of the basic data for system identification are analyzed.A mathematical model is constructed for the random response under environmental incentives.In combination with the characteristic analysis method of small signal stability,the characteristic parameters for describing the low-frequency oscillation state of the system are given.The random response data similar to noise is analyzed by the structure of the expression as system identification The rationality of the underlying data.The structure of the expression is used to analyze the rationality of noise-like random response data as the basic data for system identification.Simulation results verify the correctness and necessity of using such data for small signal stability analysis.Next,aiming at the problems that the traditional dynamic mode decomposition(DMD)does not converge to the Koopman operator spectrum and the limitation of the subspace method for the identification of nonlinear systems,an improved subspace dynamic mode decomposition(SDMD)is used for electromechanical oscillation analysis in this paper.The theoretical basis of the proposed algorithm is introduced in detail.The algorithm compresses an infinite-dimensional system into a finite-dimensional subspace for calculation.The Hankel matrix is constructed using double the amount of time-series data for orthogonal decomposition,and truncated singular value decomposition is used instead of proper orthogonal decomposition for dynamic low rank approximation.Based on this,a detailed calculation process for modal parameter identification of electromechanical oscillations by SDMD is given.Through simulation to simulate several possible situations of environmental incentives,the validity of the proposed algorithm for mode parameter extraction of power systems is preliminarily verified.Finally,the practical application of the method proposed in this paper in the analysis of electromechanical oscillations of power systems is studied.It mainly includes two aspects.On the one hand,SDMD is applied to the random data under environmental incentives to identify the electromechanical modes.SDMD can effectively overcome the problems of inaccurate identification and modal mixing of DMD,and improves the calculation efficiency compared with random subspace algorithms.On the other hand,the modal energy is calculated for the modes identified based on SDMD,and the dominant modes are extracted according to the energy order.Finally,the effectiveness of SDMD for the identification of electromechanical oscillation parameters and the extraction of the dominant oscillation mode is verified by the simulation analysis of IEEE4-machine 2-area system,IEEE 16-machine 5-area system and actual grid data.