Research On Power Quality Disturbance Detection Classification And Parameter Identification Based On Mathematical Morphology And Variational Mode Decomposition

With the further development of power systems,the widespread use of power electronic devices,the access of distributed new energy sources,the use of a large number of non-linear loads,the power quality(PQ)of power systems faces major challenges.On the other hand,users' requirements for PQ is also getting higher and higher,precision instruments are more and more sensitive to power quality,and PQ problems may cause more and more losses.Therefore,PQ problems have received more and more attention.The detection,classification and parameter identification of power quality can provide a basis for power system fault diagnosis and evaluation,and ensure the safe and stable operation of the power system.Based on the previous research,the problems of PQ disturbance detection,classification and parameter identification are discussed and researched in this paper.In the aspect of PQ disturbance detection,this paper presents a method based on morphological mixed gradient difference of basic mathematical morphological operators and an integral operator based on numerical integration and trigonometric function formula.The method combines threshold processing and logical judgment to form the final disturbance detection method.Experiments shows that the proposed method not only realizes the detection of standard signals from MATLAB and analog signal generated by Simulink,but also can detect the disturbance signal collected from the actual system,which verifies the validity of the detection method.In the aspect of PQ disturbance classification,firstly,the disturbance signal is decomposed by variational mode decomposition(VMD),then the main features of the decomposition result and the original signal are extracted by using phase shift operator(PSSO),and then the disturbances signals are classified by combining decision tree(DT).For the classification methods of machine learning such as extreme learning machine(ELM)and support vector machine(SVM),besides the features extracted by VMD combined with PSO,the feature extraction method based on correlation coefficient between signals is proposed to classify PQ disturbances signals.Finally,a comprehensive comparative analysis is investigated with other classification literature to verify the effectiveness of the feature extraction and classification method proposed in this paper.In the aspect of PQ disturbance parameter identification,five parameter identification methods are proposed to realize the identification of different parameters of different disturbance signals.Firstly,a frequency operator based on integral operator is proposed to recognize the frequency of signals in a shorter data window.Secondly,a method based on PSSO is proposed to identify the amplitudes of the swell,sag and interruption disturbance signals.Thirdly,a method based on morphological filtering and VMD is proposed to identify the parameters of transient oscillation disturbance successfully.The oscillation frequencies and attenuation coefficients of transient oscillations are identified and the identification results are compared with empirical mode decomposition(EMD).Fourthly,a new method based on morphological filtering and multiple VMD is proposed to identify the parameters of the transient oscillations mixed with flicker.The parameters are successfully identified include both the oscillation frequency and attenuation coefficients of transient oscillations and the flicker signal envolope.Fifthly,a parameter identification method based on FastICA and VMD is proposed to identify the parameters of three complex disturbances: transient oscillation mixed harmonics,transient oscillation mixed with swell and harmonics,transient oscillation mixed sag and harmonics.The identified parameters include not only the frequency and attenuation coefficient of transient oscillation,but also the amplitude and phases of harmonic signal as well as the amplitudes of the swell or sag signals.And the identificaiton results are compared with EMD algorithm and the Prony algorithm and FastICA,which highlights the effectiveness of the proposed method.