Detection Signal Processing Methods Of Wind Power Converter Based On Compressed Sensing Theory

At present,the detection signal processing methods of wind power converter are all based on the foundations of Nyquist sampling theory.These methods can produce huge sampling data,which make its storage or transmission cause “space catastrophic” consequence.Compressed sensing(CS)theory gathers signal in a lot less than the Nyquist frequency of sampling.Moreover,the signal sampling and compression makes to process simultaneously.Compressed sensing retains the key information of original signal in a certain way which is non-adaptive projection of signal,and accurately reconstructs the original signal by mathematical method.In view of the above research background,this thesis focuses on the detection signal processing methods of wind power converter based on compressed sensing theory.Not only does this method have the characteristics of compressed sampling,but have the capacity to restore and reconstruct accurately as well.So it has important research value.Firstly,compressed sensing theory is researched,simulation model of converter is built,and sparsity of the original voltage detection signal for converter is analyzed.According to the sparsity,the relationship between sparse transformation base of voltage detection signal and projection matrix is studied.By optimal projection theory,decreasing the mutuality is to optimize projection matrix.This method improves the compressed performance of projection matrix.Then,in order to solve these problems of wasted storage space and poor reconstruction performance that the three-phase voltage detection data of converter output are using CS theory directly,this thesis researches the voltage signal CS compression method of wind power converter based on coordinate transformation.By utilizing relationship of three-phase voltage,the three-phase voltage of converter output can transform one dimensional signal in terms of the coordinate transformation.One dimensional signal is compressed and reconstructed by using CS.The reconstructed signal is converted into two-phase signal.Then,two-phase signal makes coordinate inverse transformation to obtain reconstructed three-phase voltage signal.Experiments show that when using this method handles voltage detection signal of converter,it can effectively compress the original three-phase voltage data,make running time lower,have reconstruction error smaller,and save the storage space of measurement data.Secondly,the five typical greedy reconstruction algorithms are studied comparatively.Aiming at the problem such as low reconstruction performance of original signal,generalized orthogonal matching pursuit algorithm based on generalized Jaccard coefficient is researched.This method is that using generalized Jaccard coefficient criterion of similarity to replace inner product criterion of similarity is to optimize support set.Compared with the others,this researched algorithm has better reconstruction performance in same condition.Finally,considering the problem that voltage detection signal contains deep information,voltage signal reconstruction method of wind power converter is researched,which is based on intrinsic time-scale decomposition(ITD)and modified inner product reconstruction algorithm of compressed sensing.Using ITD,the voltage detection signal is decomposed into several independent proper rotation components and a residual component.However,all components are processed by using the generalized orthogonal matching pursuit algorithm based on generalized Jaccard coefficient.By reconstruction and superimposition,the original voltage detection signal of converter is obtained.This method decreases computational complexity and reconstruction error.