The Numerical Differentiation Based Eigen-sensitivity Analysis For Large-Scale Power Systems

As more and more interconnected power grids emerge around the world,the operating condition of existing power systems are becoming increasingly complicated,as a result the small-signal stability has been a high concern for power engineering.The eigen-sensitivity can quantitatively provide the trajectories of system oscillation modes when some system parameter varies,which can provide potential measurements to improve the system’s damping.Currently,the analytic eigen-sensitivity method(AES)based on the close-form sensitivity expression has been widely used.However,AES will not be applicable when the system model is inaccurate.Based on the background above,the numerical differentiation based eigen-sensitivity analysis for large-scale power system are studied.Specifically,the main works of this thesis can be summarized as follows:1)The principle of three numerical differentiation methods,the difference quotient method,the interpolation method and the modified regularization method,are elaborately investigated,and the performance of all these numerical differentiation methods are analyzed from two aspects:the effect of perturbation step setting and the effect of measurement error.2)The numerical differentiation based eigen-sensitivity analysis(NDES)calculation procedure,based on an existing commercial small signal analysis package called PSD-BPA,for a general oscillation mode analysis of large-scale power grid is then developed,which can fulfill a batch quantity calculation of some eigen-sensitivity with respect to relevant system parameters.Further,a scheme of damping improvement by adjusting related power injection is proposed.Finally,the effectiveness of NDES is validated by a simulation study on a real large-scale power system.3)To further enhance the calculation efficiency,a preliminary scheme that uses the multithreaded programming,an easy-to-use parallel computing technology in PC hardware,is developed and the resultant calculation speedup is investigated.