Neural Network Signal Processing With Nonlinear Distortions in a “Sliding Time Window”

Abstract

Continuous monitoring of the signals harmonic components level in electrical networks is an important task in ensuring high-quality power supply to consumers. This applies to both normal and emergency power system operation modes. One of the nonlinear signal distortions sources in measuring devices are nonlinear operating transformers modes. Saturation effects and hysteresis phenomena in measuring current transformers make it difficult to identify the actual operating parameters of electric power equipment. The paper shows that the apparatus of artificial neural networks can be used to control the nonlinear distortions of industrial frequency signals. The proposed algorithm based on a direct propagation neural network is tested on the example of distortion of current signals in the secondary winding of a measuring transformer. It is shown that it is possible to determine the amplitude, frequency and phase of the signal harmonic components in a “sliding time window” with an accuracy of a few percent. Estimates of the required frequency and interval of signal digitization are made; a comparison is made using the discrete Fourier transform algorithm.

Recurrent Neural Network for Controlling the Spectrum Width of a Non-Stationary Random Signal

“…Постоянный контроль уровня гармонических составляющих тока в электрических сетях также является важной задачей в обеспечении качественного энергоснабжения потребителей. Это относится как к нормальным, так и к аварийным режимам работы энергосистемы [2]. Поэтому возникает необходимость максимально быстро в режиме реального времени определить момент смены характеристик сигнала.…”