The Research On The Estimation Of Interharmonics Based On Support Vector Machine

With the development of modern power electronic technology, more and more nonlinear components were used in power system. They brought the problem of harmonic and interharmonics, which made the distortion of voltage and current more serious, reduced the power quality, especially the time-varying harmonics. People are paying more attention to them.Based on the analysis of traditional interharmonics parameters estimated, this paper takes support vector machine into interharmonics estimation, which is a new kind of machine learning method based on structural risk minimization. It makes inter- harmonic estimation into a quadratic programming problem.The support vector machine was used to estimate interharmonics parameters, through making the nonlinear relationship between sampling time and sampled values linear by kernel function; for the character of power system noise, the secondε- insensitive loss function was used to improve the estimation performance in noisy environment; iterative reweighted least squares was used to overcome the problem of computing complexity growing exponentially with the length of time series.The principles of ESPRIT algorithm was indicated by signal subspace and noise subspace which were obtained through eigenvalue decomposition of the autocorrelation matrix of sampled data. LS-ESPRIT and TLS-ESPRIT were proposed by Singular value decomposition of a HANKEL matrix which was obtained by sampled data, and they were compared with traditional implementations method by simulation. We found that TLS-ESPRIT had an impact over interharmonics frequency parameter estimation. The factors of affecting performance of TLS-ESPRIT were researched, such as parameters setting and signal characteristics.SVM was combined with TLS-ESPRIT to solve the problem of large computation and long sampling time in interharmonics estimation by using SVM alone. The TLS-ESPRIT was used to get the frequency first, and then amplitude and phase parameters were gained by SVM, the advantages of two algorithms were integrated. The simulation results indicate that this method was effective in interharmonics estimation, and it could be used in engineering. The influence of the accuracy of frequency estimation to support vector machine was researched by simulation, and the requirement of frequency estimation method combining with support vector Machine was proposed.