Research On Identification For Low Frequency Oscillation Modes Based On Robust Recursive And Adaptive Filter Algorithm

With the large units application to fast excitation system and the developmentpower grid, the widespread low-frequency oscillation seriously threatens the safeoperating of power system. The exiting on line identification algorithm is effective tosingle type signal(ambient data or ringdown data), but lacks accurate and effectiveidentification with change of the type signal(such as the disturbances occurrence insteady state). How to estimate and track the low frequency oscillation modes in powersystem during the whole process. Warning and alarming of the low frequency oscillationtimely is very important. This paper proposed that the new robust recursive and robustadaptive filtering identification method can effectively solve the above problem.Firstly, the structure and application situation of wide area measurement systemand the definition, mechanism as well as a variety of analytical methods for lowfrequency oscillation were reviewed. ARMA model for the low-frequency oscillationprinciple based on the theory of small perturbations was given, and problem ofdetermining the order of ARMA model was analysed and discussed. Combined with theactual situation of the online identification for low frequency oscillation, this paper putforward a low frequency oscillation mode identification scheme, including datapreprocessing, determination order of ARMA model, the calculation of ARMA modelparameter, the calculation of low frequency oscillation modes and the selection of thedominant modes.Secondly, this paper analysed divergence reason of conventional recursivealgorithm on ringdown data identification, and proposed a new robust recursivealgorithm based on ARMA model to ensure the algorithm under various conditions ofconvergence, by choosing the weights on the basis of the L1norm of the input-signalauto-correlation matrix and cross-correlation vector. The algorithm analysed a lot of thesimulation data from New England39bus power system, which contained steadyambient data, ambient data with perturbation and data of mode change. The resultsdemonstrated the effectiveness, robustness and fast tracking of the algorithm. Finallythe measured data of the southern power grid is also adopted to analysis and test. Theresults further validate the proposed algorithm can estimate both ambient data andringdown data.Finally, the principles and methods of low frequency oscillation identification based on least mean squares adaptive filtering algorithm was studied. This paperanalysed divergence reason of conventional least mean squares algorithm in ringdowndata, an robust least mean squares adaptive filter algorithm was applied to estimate thelow frequency oscillation modes. The algorithm analysed the simulation data from NewEangland39bus power system and the measured data in China southern power grid, theresult validated robustness and validity of the proposed algorithm, which can not onlyestimate both ambient data and ringdown data, but also eliminate the effects of outlierdata on identification.