Study On Detection And Identification Of Power Quality Disturbances

The quality of electric power has seriously affected the safety and economic operation of the whole power system,so it is very important to improve and ameliorate the quality of power.With the increase of nonlinear and impulsive load in power system,the distortion of voltage and current waveform in power gridis becoming more and more serious,resulting in deterioration of power quality.Power quality detection and identification is an important content of power quality monitoring.It is of great significance for the optimization of power quality problems to detect the disturbance events quickly and efficiently and to accurately locate and classify them.The work of the research on the detection and identification of power quality disturbance is as follows:Power quality disturbances mainly consist of two categories,steady state and transient state,and the related standards put forward more detailed classification basis,which show that different types of disturbance to show the time-frequency information.With the extensive application of a large number of power electronic devices,the signal of power quality is becoming more and more non-stationary and nonlinear.The traditional method of power quality detection is not suitable for nonlinear and non-stationary signal analysis.For effectively analyze the quality of electric energy,some common disturbance signals has build by the mathematical model,and the empirical mode decomposition(EMD)and the local mean decomposition(LMD)to simulate for them.An improved LMD detection algorithm is proposed based on the original signal to simulate the actual signal so as to get the detection result more quickly and accurately.And the simulation analysis proves the validity of the method.A method combining the improved LMD with entropy is proposed to extract the feature of the power quality disturbance signal,so as to classify the power quality disturbance and provide characteristic vectors.In this paper,several types of classification recognition methods are studied.The support vector machine(SVM)is used to classify and identify the sample set,and the results of classification recognition are analyzed and improved.At the same time,the SVM parameters are optimized by the global optimization of genetic algorithm,and the final classification results are obtained.The classification accuracy rate can reach 98% and above through the analysis and processing of the sample data,which can completely distinguish and identify all kinds of disturbance signals.Finally,the validity of the waveform data verification algorithm based on Simulink simulation platform and the micro-grid system experiment platform with distributed power supply is presented.