Research On Disturbances Detection And Classification For Power Quality Of Power System

Electrical energy is applied to various fields of production and life,which is clean and environmental protection,and it is easy to be controlled and transformed.The power quality not only affects the interests of users,but also affects the safe operation of power grid,therefore,the detection of power quality problem has important significance.The main power quality problems are classified,and the causes and hazards of various types of power quality problems are analyzed.According to the common characteristics of transient and steady-state power quality disturbance problem,the mathematical models of the disturbance are established and the research status of power quality disturbance detection methods and classification methods are analyzed.In this paper,the main work content is as follows:The Wavelet Packet Transform is used to analyze the power quality disturbance signal,and the wavelet packet energy distribution is disposed to a new wavelet packet energy distribution with marketed,the sampling frequency,wavelet basis function and decomposition level of Wavelet Packet Transform are selected according to the characteristics of the disturbance signal.The energy distribution of harmonic disturbance and transient oscillation disturbance in corresponding nodes of new wavelet packet energy distribution is bigger than the other power quality disturbance problems,therefore,the energy of corresponding node in the new wavelet packet can be extracted as characteristic vector of these disturbances,and the feasibility of this method which using wavelet packet transform to extract disturbance characteristic vector can be analyzed through the simulation.The HHT transform is adopted to analyze various types of power quality disturbance signals,and the instantaneous frequency of HHT analysis result is used to estimate start-stop moment of different kinds of disturbance,the amplitude andfrequency characteristics of disturbance signal can be extracted according to the instantaneous amplitude and marginal spectrum.However,the feature extraction effect of HHT transform is not obvious.In order to find a more suitable method,the S transformation is used to analyze the various types of power quality disturbance signals.The change of highest frequency amplitude is used to estimate the start-stop moment of disturbance,the amplitude and frequency characteristics of disturbance signals can be extracted from the fundamental frequency amplitude changes and time average amplitude sum of squares.By comparing the effect of the two methods,it can be seen that the estimation accuracy of start-stop moment is higher when using S transformation,and the characteristic changes obviously,convenient for threshold value select,and easier to implement.For the single characteristic vector can't represent the differences of all the power quality disturbance signal effectively,the characteristics of different feature detecting methods are analyzed,and the multiple features' combined logic of the power quality disturbance is researched in this paper.The signal with harmonic and transient shock can be classified according to the obvious differences of the harmonic signal in the wavelet packet energy distribution and the obvious frequency characteristic in S transform,and the other amplitude characteristics and singularity characteristics can be classified by using S transform.The multi-feature combination classifier is researched by using multiple probabilistic neural network,and simulation results demonstrate that the average classification accuracy can reach99.25% when the noise intensity is 40 dB,and the average classification accuracy can reach 88.38% when the noise intensity is 40 dB.