Research On Low Frequency Oscillation Analysis Based On Wide Area Measurement System

Low frequency oscillation seriously threatens the safe and stable operation of power system,thus,it is one of the important topics in the field of power system research.Generalized low frequency oscillation can be divided into ultra-low frequency oscillation and low frequency oscillation according to frequency.The oscillation frequency of low frequency oscillation and ultra-low frequency oscillation are 0.1 Hz to 2.5Hz and less than 0.1 Hz respectively.Furtherly,low frequency oscillation can be divided into free oscillation and forced oscillation according to mechanism.Due to inaccuracy of load modeling caused by timely variation,the analysis result of free oscallation is inaccurate.The reason of forced oscillation is that there is periodic oscillation source in the system.Thus,a key problem of forced oscillation is how to locate the forced oscillation source.As for ultra-low frequency oscillation,there is a lack of fast and effective method to analyse the stability contribution of the unit in ultra-low frequency band based on measured data.Based on aforemention problems,the main contributions of this dissertation are as follows:1.An on line load modeling method based on subspace method suitable for free oscillation analysis is proposed in this paper.Firstly,small disturbance data which always exists is selected as identification signal and state space model of load is identified by subspace method.On line load modeling is accomplished which solves the inaccuracy problem caused by timely load variation.Since the state space model is not unique,an unknown linear transformation exists between state space load model identified by subspace method and real state space model of load.It is proved that the unknown linear transformation has no influence on the result of small signal stability evaluation through analytical approach.2.The mechanism of load modelling under small disturbance condition has been further explored.It is proved that the individual load to be identified and the rest of the system forms a closed-loop system.The impacts of disturbances entering the feedforward(internal disturbances)and feedback channel(external disturbances)are examined analytically.It is found out that load identification is closed loop identification and general external disturbances function as excitation signals thus help improve load modelling accuracy,while internal disturbances play a role in the opposite direction.Particularly,the impact of closed loop identification on load identification can not be neglected under small disturbance condition.Closed loop identification can be solved by prediction error method(PEM),provided that some mild conditions are met These conditions are discussed and implementation of PEM based on a Kalman filtering formulation is detailed.3.A method to locate oscillation source is proposed based on power spectral density.The relationship between the power spectral density of system outputs and transfer function from assumed disturbance to system outputs is derived.The location of the oscillation source can be identified by testing whether or not the relationship holds true for each generator of a power system.4.A method is proposed to locate the forced oscillation source,which is often resulted from a continuous disturbance on unit's mechanical power.By judiciously designing a Luenberger observer using an eigenstructure assignment,the relationship between outputs of observer and disturbance in mechanical power of a unit is decoupled.The introduced observer cannot only locate the forced oscillation source in mechanical power,but also estimates the amplitude of the oscillation source.5.A method for analyzing the stability contribution of units in ultra-low frequency band is proposed.The single-frequency system model used for the ultra low frequency oscillation analysis is described firstly.Then,the real part of the transfer function of governor-turbine system at frequency of ultra-low frequency oscillation can reflect the stability contribution of units in ultra-low frequency band which is figured out based on Nyquist criteria.Finally,the detail about how to identify the real part of the transfer function of governor-turbine system at frequency of ultra-low frequency oscillation are given based on power spectral density and the fact that data during oscillation for identification can guarantee reliability is pointed out in this paper based on closed loop characteristic of the single-frequency system model.