Harmonic Analysis Approach In Power System Based On Neural Network Algorithm

In recent years, the harmonic pollution of the power system is getting more and more serious with the far-ranging of power electronics technologies. It has become a curse to power system, which deteriorates the power quality. Harmonic measurement is the important part of harmonic issue and harmonic measurement method is the key of the harmonic measurement. This paper primarily studies the harmonic measurement methods.The paper amply introduced the principles,application situation and existed problems of several harmonic analysis algorithms which are used frequently. In the view of the several problems of different algorithms such as the low computation accuracy and the big computation quantity and so on, an artificial neural network model based on the Fourier basis functions was established in the paper with the harmonic model of the electric power system. An artificial neural network algorithm based on the Fourier basis functions and improved Prony algorithms which based on the gradient descent rule and least squares method were researched. To validate the astringency of the algorithm, the convergence theorem of the algorithm was proposed and proved. The theory gist to select learning rate was provided by the convergence theorem. To validate the validity of the algorithm in the field of the power system harmonic analysis, the computer simulation and other algorithms comparison research were given in the paper. The research results showed that the harmonic analysis approach of the power system proposed in the paper had the high computation accuracy and the fast computation speed. Especially, the improved algorithm based on an artificial neural network and Prony algorithm didn't deal with the operation of inversion matrix in traditional Prony algorithm, improving the practicality value of the Prony algorithm.Certainly, the improved algorithm based on an artificial neural network and Prony algorithm studied in the paper had some limitations, especially when noise pollution, the computation accuracy of the amplitudes and phases can't obtain high precision, so waiting for further research.