Research On Characteristics Extraction And Analysis Method Of Electromechanical Oscillation For Interarea Power System Based On Optimized Dynamic Mode Decomposition

With the rapid construction of UHV power grid and continuous expansion of power grid,the dynamic characteristics of the power system is becoming more and more complicated,low-frequency oscillation shows obviously spatial and temporal distribution characteristics,which bring lots of serious challenges to the modes extraction and analysis of low-frequency oscillation.In recent years the wide area measurement system which with the help of phasor measurement unit has been rapidly developed and widely used in large-scale interconnected power grid.The identification of electromechanical oscillation parameters is mainly based on WAMS,the parameters of oscillation frequency,damping ratio and oscillation mode can be extracted by the algorithms of signal analysis or parameter identification which is of great significance to extract and analyze electromechanical oscillation mode for the safety assessment of power system.In this paper,a new method of electromechanical oscillation parameter identification and analysis is proposed based on wide-area measured data.Dynamic mode decomposition algorithm(DMD)was used to analyze the global behavior of electromechanical oscillation.The core objective of dynamic mode decomposition algorithm is to find the low dimensional approximate matrix.We hope to get a low dimensional approximate matrix whose eigenvalues and the corresponding eigenvector could be employed to describe the oscillation features of the state space matrix.At this point,the electromechanical oscillation modes would be extracted under the premise of the state space matrix unknown.In the process of data collection from actual system,noise pollution,which brings a great difficulty to mode extraction,is inevitable because of the measurement environment and the error of measurement devices.However,the measurement noise has not been given sufficient consideration in the dynamic mode decomposition algorithm,in this paper,the optimized dynamic mode decomposition(OpDMD)algorithm that is robust against noise is introduced to extract the oscillations mode parameters and select the dominant modes from the multi-channel synchronous measurement data sets.To reduce the effect of noise on the estimation results,we employ the finite difference style approximation based DMD to obtain the initial eigenvalues firstly.On the basis of the initial eigenvalues,the goal is to minimize the difference between the reconstructed data and the actual data,the variable projection is use to find the eigenvalues and corresponding eigenvectors,so the optimized dynamic mode decomposition can extract obvious spatial and temporal distribution characteristics of the global oscillation from the different noise intensity of WAMS directly,at the same time,the modal energy obtained along with the eigenvectors is a valid quantitative indicator to select the dominant modes.As can be seen from the simulation analysis of simplified 14-generator model of the SE Australian power system,16-machine,68-bus system and actual measurement data,by employing the optimized dynamic mode decomposition,we extract the eigenvalues and the corresponding dynamic mode which provide oscillation temporal characteristics(oscillation frequency and damping ratio)and spatial characteristics(mode shape)from the multivariate coupled disturbance signal of the power system,at the same time,the dominant modes can be picked up.