In order to solve the problems of high-speed transmission

large-capacity storage for a long time

and low reconstruction performance on wind power converter voltage detection signal

we proposed a converter voltage signal reconstruction method of combining the intrinsic time scale decomposition（ITD） and reconstruction algorithm of compressed sensing based on improved inner product.This method adopts the simultaneous operation of voltage signal compressing and sampling. Firstly

the voltage signal was adaptively decomposed into several proper rotation components and a residual component by ITD. Sparse representation of all components was used by the discrete cosine transform. The underdetermined equation based on the Gauss random matrix was established. Then

each component of the voltage signal was reconstructed by a generalized orthogonal matching pursuit algorithm of compressed sensing

which used the proposed matching measurement criterion based on general Jaccard coefficient instead of inner product standards to optimize the support set. Finally

all reconstructed component signals were combined into getting the original converter voltage detection signal. Simulation shows that this method in which the ITD and reconstruction algorithm based on an improved inner product of compressed sensing are combined can not only improve the reconstruction accuracy of converter voltage detection signal and reduce the reconstruction computational complexity

but also enhance the operational efficiency.