Research On Power System Dynamic Parameter Estimation And Model Reduction

With the widely application of the new type power electronic equipment and the gradual formation of the ultra-high voltage interconnected power grid,the power system dynamic models become complex and its scales become large,which brings new challenges to the security and stability of the power system operation.Aiming at the new characteristics of the power system models,this work studies the power system dynamic parameter estimation and model reduction based on the wide area measurement system from two aspects,the accurate modeling of the power system and the efficient application of the dynamic model.In aspect of the accurate modeling of the power system,this work takes the synchronous generation system as the research object,and studies the grid disturbance based online parameter estimation methods from three directions including discontinuous model identification,multi-sce-nario identification and small-disturbance identification.Firstly,in the view of the discontinuous dynamic behavior of the dynamic components,a space regularization based parameter estimation method is proposed.The dynamic behavior of different state spaces is equivalent to a unified form by using the Heavyside function.Then non-linear programming algorithms solve the discretized equivalent dynamic equation constrained op-timization problem,to achieve the accurate and efficient identification of discontinuous dynamic systems.In addition,a two staged parameter estimation method is proposed for identifying the parameters of the discontinuous switchings,which is able to improve the corresponding estima-tion accuracy.Secondly,in order to improve the results of the online parameter estimation,this work pro-poses a systematic method for the multi-scenario parameter estimation.This method can be im-plemented through four crucial steps.The first step is to select the identifiable parameters via the multi-scenario identifiability analysis.According to the scenario identifiability index,the second step is to rank the scenarios and select dominant ones.This approach is able to maintain a balance between accuracy and efficiency of the multi-scenario parameter estimation.After the precondi-tioning of parameters and scenarios,the third step is to utilize the scenario-decomposition based reduced-space interior point algorithm.This algorithm handles the scenario associated matrix op-erations in different processors,to realize an efficient solution of parameter estimation.The forth step is the bad scenario detection and identification.Chi-square test is applied to eliminate the bad scenario and improve the estimation results.Thirdly,in the view of the situation that under small disturbances the measurement error is compatible with the disturbance variations,this work proposes a parameter estimation method considering input errors,which can correct the influence of input errors on estimation results ef-fectively.Simultaneously,a Lagrange multiplier based input variable selection method is pro-posed to reduce the degrees of freedom of optimization problem,improve the computational effi-ciency and avoid the over-fitting phenomenon.What's more,the Tikhonov regularization method is implemented via introducing the prior knowledge of parameters to enhance the parameter iden-tifiability.Then the optimal parameters can be selected via the cross validation approach to ensure the parameter estimation accuracy.In aspect of the efficient application of the dynamic model,the most time-consuming opera-tion of extended Krylov subspace method for solving the Lyapunov equation is analyzed.Accord-ing to the sparse structure of the Jacobian matrix of the power system dynamic model and the dual characteristics of the Lyapunov equations,a sparse extended Krylov subspace method is proposed for the power system dynamic model reduction.This method can realize a fast and accurate dy-namic model reduction of large-scale power systems.The failure reasons of balance truncation in the unstable system are also analyzed,and the proposed reduction method is extended for unstable systems by using the alpha shift approach.Furthermore,a PSS parameter optimization method based on reduced order model is proposed.Instead of using the original model,the PSS parameters are tuned via the reduced model efficiently,and the optimized PSS controller can suppress the low frequency oscillations in power systems.