Intelligent Sensing Of Power System Low Frequency Oscillations

In the context of the expansion of power system interconnection,low frequency oscillation(LFO)is still one of the key problems threatening the power systems.Low frequency oscillations in power systems usually fall into two types,i.e.,natural oscillations and forced oscillations.Due to the different suppression or elimination measures for different types of LFO,the operators of power system must identify LFO type accurately and quickly.If the identification result is forced oscillation,it is necessary to locate the disturbance source in time to eliminate the oscillations.The large-scale deployment of phasor measurement unit(PMU)and the rise of artificial intelligence technology provide a new method to solve the problem of LFO perceiving.This paper first summarizes the theoretical basis of LFO and analyzes its characteristics,and then reveals the limitations of traditional LFO analysis methods applied to modern power system and the applicability of AI methods to solve the problem of LFO.Then,aiming at two key problems in the field of LFO perceiving: LFO type identification and forced oscillation disturbance source localization,an identification method of LFO type based on multi-dimensional features and Relief F-m RMR and a hierarchal localization method for FO disturbance source based on deep transfer learning(DTL)are proposed.The main contributions of the thesis are as follows:(1)The mechanisms of natural oscillations and forced oscillations are deduced based on mathematical models for power systems,and the characteristics of two types of oscillation are analyzed.On this foundation,aiming at the two key problems of LFO analysis: LFO type identification and forced oscillation disturbance source localization,this paper illustrates the limitations of traditional methods and the applicability of artificial intelligence methods in solving these two problems.At last,the theoretical analysis is verified in the simulation.(2)An identification method of LFO type based on multi-dimensional features and Relief F-m RMR is proposed.Firstly,the multi-dimensional feature index set is constructed to comprehensively describe the characteristics of LFO,which includes time domain,frequency domain,energy,correlation,complexity and mode.Then,Relief F-m RMR is utilized to select features from LFO index set,which combines the high efficiency of Relief F and the ability to reduce feature redundancy of m RMR.At last,a modified support vector machine with genetic algorithm(GA-SVM)optimizing the key parameters is used to obtain the model to realize the LFO type identification.Finally,the good robustness,noise immunity and practicability of the proposed method are verified in the simulation system and the actual system.(3)A localization method for FO disturbance source based on DTL in large-scale interconnection of power systems is proposed.The localization process is decomposed into system-level localization and area-level localization,which relieves the data communication pressure and computation burden.By adopting principle component analysis to extract the most representative information at system-level and smooth pseudo Wigner-Ville distribution(SPWVD)to characterize FOs signals,graphical representations of FOs at system-level and area-level are obtained respectively,transforming the localization problem to image recognition problem.Finally,a two-stage DTL algorithm is developed to locate FOs disturbance source.In the case studies,the impact of different inputting features,noise and topology change on the proposed method is further explored,which demonstrates the high accuracy,good anti-noise performance and robustness of the proposed method.