Research On Oscillation Source Location Of Power System Forced Oscillations

With the interconnection of large-scale power grids,the increasing size and complexity of power grids,the adoption of ultra high voltage AC/DC transmission,the connection of various renewable generations and power electronic equipments to power grids,the nationwide resource allocation of power supply has been optimized,which greatly improve the economy and reliability of power grid.Meanwhile,however,the stability of power systems is becoming an increasingly noticeable problem,threatening the safe and stable operation of power systems.When periodic disturbances occur in the prime movers and governors and excitation systems of generators,load variations and other devices in the large-scale power grids,power system forced oscillations will be triggered.Especially when the disturbance frequencies become close to the natural frequencies of the power system oscillation modes with a low damping ratio,resonance effect occurs,which will magnify the oscillation source many times and arouse large-magnitude forced oscillations.The case where the damping ratio of the mode is exactly 0 will result in serious resonance effect and cause the oscillation amplitude to reach maximum.With regard to power system forced oscillation,this paper proposes several methods to locate forced oscllation source to generators and the control devices of generators.The main work is as follows:(1)A location method is proposed which uses unknown input observer(UIO)-based method to isolate oscillation source and accordingly,realize oscillation source location at device level.Firstly,a linear power system model is derived.Typical oscillation sources,namely,oscillation sources from prime movers and speed governors,excitation systems,are included in the model as unknown inputs.Then,a systematic method using UIOs and fault isolation scheme to locate oscillation source is proposed.Using the disparities between the model responses generated by UIOs and the recorded responses,the proposed method establishes an one-to-one relationship between a residual and an oscillation source.The location of oscillation source at device level is accomplished on the basis of the different sensitivities of residuals to oscillation sources from different devices.The detectability of oscillation source and the numerical check are also discussed.The simulation examples show that the location of oscillation source by the proposed method is effective and is accurate to the control device that the oscillation source originates from.(2)Considering that forced oscillations are basically caused by oscillation sources from generators,a method for locating oscillation source which is based on the sensitivity design of observer is proposed.On basis of the H_-/H_?norm of the transfer function matrix,the inequality constraints are given which assign different sensitivities to oscillation sources inside the generator and outside the generator according to the generalized KYP lemma.Using linear matrix inequality(LMI)to search and optimize the feedback matrix of the observer,the observer residual is designed to be sensitive to the oscillation sources inside the generator while it remains insensitive to the oscillation source from the outside power grid(decoupling).The oscillation source generator could be located according to the spectrums of the observer residuals of different generators.Demonstrative examples of a 39-bus system and the 547-machine 8647-bus North China Power Grid show that the proposed method is effective,intuitive and robust.(3)The location of the control device that produces oscillation sources is crucial to the elimination of forced oscillations.A method to locate oscillation source to generator control devices is proposed which is based on power spectral density prediction of generator model responses.The notion“model response of generator”is proposed and derived,based on which the method of power spectral density prediction for generator model responses is developed.For non-source generators,the power spectral density of responses is determined by the generator terminal voltage(or generator current).The generator with distinct mismatch between predicted and measured power spectral density is identified as the oscillation source generator.The forced components of different responses are affected by oscillation sources from different control devices to different extents.It further enables the proposed method to identify the control device producing the oscillation source.Demonstrative examples of a 39-bus system and the 547-machine 8647-bus North China Power Grid show that the proposed method is effective and intuitive.(4)Periodic disturbances in prime movers and governors(mechanical parts)are major causes for forced oscillations.Based on the Phillips-Heffron model of multi-machine systems,the transfer function matrix from disturbances in mechanical parts to mechanical powers is derived,which is proven to inherently have the parameter characteristics of diagonal dominance with parameters in the practical ranges of real power systems.According to the property,the mechanical part of the generator whose mechanical power has the largest power spectral density is the possible oscillation source for forced oscillations caused by periodic disturbances,malfunctions and parameter abnormities in mechanical parts.Meanwhile,the influences of periodic disturbances in loads and exciters on mechanical powers are considered.The transfer function matrices from loads and exciters to mechanical powers are also derived,showing that the influences of load fluctuations and excitation disturbances on mechanical powers are generally very small compared with mechanical disturbances.As a result,it is reasonable to identify the generator with the largest power spectral density of mechanical power as oscillation source.Demonstrative examples of 39-bus system and North China Power Grid show the effectiveness and simplicity of the proposed location method.(5)A data-driven method to locate oscillation source is proposed which uses the spectral Granger causality analysis and arrival delay estimation of forced oscillation.The basic idea of Granger causality and the theory of the spectral Granger causality is provided.The construction of the AR models and the feature selection based on principal component analysis in causality analysis are given.The application of the spectral Granger causality analysis to locate forced oscillation source is proposed.Moreover,to provide comparison and confirmation for the location results,a method of arrival delay estimation of forced oscillation is proposed to determine the arrival sequence of forced oscillation at different generators.It then helps to distinguish source generator from other generators.Cross correlation function is adopted to estimate time delays of oscillation arrivals and it identifies the generator which oscillates first as the oscillation source.The effectiveness of the proposed methods is validated by study cases.It is shown that the proposed method is model-free and needs only few recorded data.