Resarch On Harmonic Current Detection And Suppression Technology Based On RBF Neural Network

Harmonics not only affect the power quality,but also reduce the safety and stability of the power grid.With the progress of technology,a large number of new power electronic products containing nonlinear components are connected to the power grid,and the harmonic pollution of power system is becoming more and more serious.Meanwhile,the application of new grid-connected technologies,distributed generation technologies and the acceleration of smart grid construction have also posed new challenges to the traditional power system.Harmonic detection and harmonic suppression technology is the research direction of this paper.Based on the analysis of the current research status at home and abroad,we have deeply studied the active power filter（APF）harmonic suppression method based on RBF neural network prediction.The main contents of this paper are as follows:Research on harmonic detection method of power system:The p-q harmonic detection and i_p-i_q harmonic detection based on instantaneous reactive power theory are deeply studied,and the two methods are modeled and simulated by MATLAB-Simulink.It can be verified by simulation that both methods can accurately detect the harmonic current in the power network when the voltage source does not produce distortion.When the voltage source is distorted（excluding other external factors）,the harmonic detected by p-q method does not match the actual harmonic current seriously,while i_p-i_q method can still detect the harmonic current accurately.Research on harmonic suppression method of power system:Harmonic suppression methods in power network system are divided into traditional and passive harmonic suppression.Through analysis,it is known that the traditional suppression technology has high cost and low efficiency,and usually needs to introduce passive suppression methods to improve the performance of harmonic suppression.On the other hand,passive filter and active filter are two commonly used passive harmonic suppression methods.Through theoretical analysis and MATLAB-Simulink modeling and simulation,it is found that passive filter can suppress harmonic to a certain extent,but its filter performance is susceptible to changes in power network parameters and component parameters.Research on APF harmonic suppression based on RBF neural network prediction:By analyzing the working principle,mathematical model and control method of APF,the problem of time delay in the working system is found.Therefore,the RBF neural network prediction algorithm will be used to solve this problem.First,the gradient descent method and K-means clustering method are used to calculate and optimize the base width,weight and cluster center of RBF neural network.Secondly,select the appropriate samples and train the RBF neural network offline so that it has the best prediction performance.Then,the whole power system is simulated,and the trained neural network is used for harmonic current prediction.The simulation results show that the method we used can accurately predict the reference value of the harmonic current at the next moment,and the command signal drives the compensated current generated by APF according to the reference value,which can eliminate the time delay and ultimately achieve the purpose of real-time harmonic suppression.