Research On Detection And Recognition Algorithm Of Multiple Power Quality Disturbances

Power quality is an important indicator of whether the power supply system is normal,due to the grid connection of new energy generation and the access of various non-linear loads,the power quality problem becomes more serious and complicated.In order to improve the quality of electrical energy,different types of power quality disturbances(PQD)need to be classified and detected before improvement.The power quality disturbances that exist in a real power system are not consist of only one kind in the same time period,but are caused by the superposition of many kinds of disturbances on the signal waveform.so researching these signal waveform detection algorithms has lots of significance for improvement.Aiming at the problem of disturbances types and the difficulty of identification,firstly,this thesis analyzed the definition of disturbances and used the improved wavelet algorithm to solve the problem of a large amount of redundant data and noise in the signal waveform.Due to the time-frequency accuracy of the S transform cannot meet the requirements of composite disturbances,this thesis proposed frequency divided S transform(FDST)detection algorithm.Then the simulation and comparative analysis are carried out.Finally,a neural network based on Radial Basis Function(RBF)was designed to identify the type of disturbance,and a Particle Swarm Optimization(PSO)algorithm was used to determine the network parameters.The identification of the test data samples verified that the algorithm based on the frequency divided S transform is more suitable for the classification of multiple disturbances.The main research contents are as follows:(1)Studied the classification and definition of power quality.The mathematical characteristics of different single perturbation signal waveforms are analyzed in detail,and the mathematical expression of multi-perturbation superposition is constructed.At the same time,in order to preprocess the data collected by the power system,the commonly used thresholds have been improved again,and the effect of data denoising has been verified.(2)The application principle of signal transformation based on S transform is studied.After processing the signal through the S transform algorithm,various parameters reflecting its characteristics can be obtained.Modifying the window width factor can improve the time-frequency performance of the algorithm.(3)The influence of different window width adjustment coefficients on the time frequency resolution of the detection algorithm is studied,and an improved S-transform processing method which is FDST is proposed,and the derivation process is given.The simulation results show that compared with the traditional phase-shift detection method,the detection accuracy of the start and end time,amplitude,and frequency of the disturbance signal is higher,which is more conducive to the extraction of disturbance feature quantities.(4)The principle of RBF neural network is studied.The features which is used to classify power quality disturbances was extracted through FDST.The local response characteristics of the RBF neural network greatly improved the speed of network training.At the same time,PSO algorithm was used to optimize the calculation of some parameters in the network.Then compared with the recognition accuracy before optimized.The simulation results show that the detection and recognition algorithm based on the FDST and RBF neural network has high classification accuracy,and the results are almost the same under different noises,which is more suitable for the analysis of real power quality disturbances.