Research On Power Quality Disturbance Classification Based On Extreme Learning Machine

Correctly identifying the category of power quality disturbance is the premise of taking corresponding control and compensation measures in time.The kernel extreme learning machine has good stability,generalization ability and has the characteristics of simple structure and fast running speed,which is very suitable for practical application in engineering.The dissertation studies the kernel extreme learning machine in detail.The classification method of limit learning machine based on particle swarm optimization(PSO)algorithm is proposed to optimize kernel function,which is applied to the classification of power quality disturbance,and a good classification effect is obtained.The dissertation analyzes the application of the extreme learning machine in the classification of power quality disturbances in detail.According to the characteristics of different generalization ability under different parameters,the characteristics of the kernel extreme learning machine are analyzed,including mononuclear function and mixed kernel function,in which the mononuclear function contains radial base kernel function and polynomial kernel function,and the mixed kernel function contains radial basis polynomial kernel function and radial basis linear kernel function.Simulation results show that the mixed kernel extreme learning machine achieves better classification results.In order to obtain the optimal kernel parameters to improve the classification accuracy,the particle swarm algorithm is used to optimize the kernel parameters of the kernel extreme learning machine.The output weight is determined by the least squares algorithm to form a kernel limit learning machine based on particle swarm optimization method.Considering the influence of particle swarm inertial factor on search ability,the inertial factor is analyzed.The particle swarm optimization algorithm is used to optimize the RBF kernel function parameters,regularization coefficients,and the mixtures of kernel function parameters.The simulation results show that the mixtures of kernel extreme learning machine optimized by particle swarm has higher classification accuracy than the extreme learner and the mononuclear kernel extreme machine.The PSO-KELM is used for the classification of power quality disturbance signal,and the particle swarm algorithm optimizes the mixed kernel parameters of the kernel extreme learning machine.By using the difference of wavelet coefficient energy contained in each frequency band,the characteristic table of power quality disturbance signal is established,which contains the characteristic energy value range of power quality disturbance signal given by a large number of wavelet packet decomposition.The simulation results show that the classification accuracy of the particle swarm optimization radial basis polynomial hybrid extreme learning machine is higher than that of the extreme learning machine and the particle swarm optimization the mononuclear kernel extreme learning machine.