Research On The Algorithm Of Intelligent Detection And Suppression Of Harmonics In Microgrid

Micro grid is a small power system with strong flexibility containing distributed generation system,power accumulation and load.It can promote the energy utilization and reduce the loss of transmission.It is by so far an effective way for the power shortage.With more and more nonlinear elements appear in the power gird,the amount of harmonic wave is rising,the reliability and the utilization of power is affected.Therefore,it requires to analysis the harmonic to optimize grid structure during designing in order to reduce the amount of harmonic occurred by undesirable structure of micro gird.It also requires a high efficient and reliable harmonic restraining algorithm to compress the harmonic.This essay will aim at two parts:the harmonic detection and the suppression.Aiming at the disadvantage of low calculation speed and accuracy,this essay puts forward a new intellectual algorithm for integer harmonic detection and inter harmonic detection.In the detection of integer harmonic,the driving functions are changed into sine and cosine and the calculation in hidden layer is moved to the input layer.In the detection of inter harmonic,by using the feature of the same angular frequency in the paired sine and cosine group,this essay formed a paired neural network structure,and use variable step size algorithm to speed up the convergence of error.When the harmonic signal comes,the customized neural network samples a certain amount of time-current(or voltage)value as the sample book for network’s training and equating.After the equating,the frequency,the amplitude and the phase of harmonic can be calculated from the weight and the offset matrix.Through the simulation,the new integer detection algorithm can reach the same training accuracy of the traditional one with 4%duration of training.The new inter harmonic detection algorithm can reach 4times the accuracy of traditional one with 38%duration of training.In the part of suppression,to rise the accuracy of command signal,this essay design a scheme combining neural detection algorithm and the i_p-i_qalgorithm in parallel.When the harmonic is entering or the information of the harmonic is changing,the neural network samples and equates the harmonic signal,use the i_p-i_qalgorithm to form the command signal when the calculation is running and use the information from the neural network to form the command signal when calculation is done,therefore,the accuracy of command signal and the suppression effect are improved.Aiming at the command signal tracking algorithm’s disadvantages of low control effect and high switching loss,this essay puts forward a loop control algorithm based on fuzzy control.The fuzzy controller takes difference between the gradient of instructed and the actual signal and the second derivative of the instruct signal as input signals.According the tendency of command signal and actual signal and correction for the approximate formula of stabilized-frequency algorithm,this essay puts forward a new set of control regulations and optimized membership curves for both input and output signals under the thought of minimize the fluctuation of transistors frequency to reduce the total harmonic distortion as well as the frequency of transistors.Through the simulation,comparing the new intellectual suppression algorithm and the traditional algorithm,the frequency of switch reduces 3.4%and THD reduces 2.3%.