Wind Farm Harmonic Detection Based On Improved Wavelet Transform And Neural Network

According to the structure of the wind turbine grid-connected system and the equivalent circuit,the mathematical model of the wind turbine grid-connected system is derived.The simulation model of each subsystem MATLAB/Simulink is established.The output voltage and current of the wind turbine grid-connected system are analyzed.The harmonic characteristics lay the foundation for harmonic detection and optimal control.A harmonic detection method for wind turbine grid-connected system based on improved wavelet transform is proposed.Comprehensive analysis and comparison of analog filter method,FFT method based on frequency domain analysis,wavelet transform method,active separation method based on Fryze time domain analysis and power system harmonic voltage detection method based on neural network,in order to further improve harmonic detection accuracy And the rapidity,a harmonic voltage detection method for wind turbine grid-connected system with improved wavelet transform is proposed.Firstly,the voltage signal of the wind power grid-connected system is divided intolow-frequency segment and high-frequency segment by db20 discrete wavelet transform.The low-frequency segment signal is decomposed by wavelet multi-scale algorithm,and the high-frequency segment signal is decomposed by wavelet packet.The fundamental wave and each harmonic component in the signal;then,the harmonics of the wind power grid-connected power system are detected by effectively extracting the harmonic components of the specific frequency segment for reconstruction.The harmonic detection optimization method of the rough grid PCA-Elman neural network wind turbine grid-connected system is proposed.The Elman dynamic neural network was introduced to establish the harmonic detection model of the wind turbine grid-connected system.The principal component analysis(PCA)was used to extract the features of the harmonic data to optimize the input of the neural network,and the excitation function and network structure were improved to seek the function convergence speed.And the optimal solution of detection accuracy;for the Elman neural network detection model,there is a large detection error and fluctuation of detection accuracy at the peak of wind speed fluctuation.It is proposed to use the rough value theory to correct and compensate the detection value to further improve the detection accuracy.The simulation research on the grid-connected system of 1.5MWdoubly-fed induction wind turbine is carried out.The results show that compared with the neural network harmonic detection method,the improved harmonic detection method of wavelet transform can realize the rapid and effective detection.The voltage harmonic component of the wind turbine grid-connected system improves the accuracy and speed of harmonic detection.