Separação e estimação de harmônicos e inter-harmônicos em sinais do sistema elétrico de potência baseadas na análise de componentes independentes

Abstract

The Electric Power System (SEP) is increasingly suffering from electromagnetic disturbances that impact system consumers at their various levels of supply. The constant growth of distributed and renewable generation contributes to the need to revise the limits established by regulatory standards. Voltage sags, Flicker, phase unbalance and harmonic distortion are among the events that can deteriorate the quality of electrical energy (QEE). In this context, this work presents a new method for harmonic and interharmonic separation in SEP signals based on Independent Component Analysis (ICA). The first step of the method is to assemble a mixture matrix by the Single Channel Independent Component Analysis (SCICA) method with M late versions of the analyzed signal, and then apply the second order blind identification algorithm (SOBI) for estimation of independent components. Next, the signals are grouped through matched filters tuned to the harmonic frequencies of interest. Adaptive FIR filters are automatically designed to filter each harmonic based on the response of the harmonic components detected by the matched filters. Interharmonic component pairs are identified by spectral similarity after deletion of unwanted components and are extracted from the signal. To evaluate the performance of the method, synthetic and real signals are applied, with harmonics, interharmonics, and simultaneous disturbances. The results are evaluated by the relative mean square error (RRMSE) and relative spectral difference (RSD), comparing the method outputs with the original signals. The method showed good performance in the extraction of independent components, even in the presence of time varying harmonics, low signal to noise ratio and in the presence of other disturbances. Results were obtained with the lowest RRMSE in simulated cases with different sag levels when compared to other estimation methods and harmonic separation. Actual voltage and current signals were applied to the method, and numerous harmonic components were extracted, which allowed the calculation of signal harmonic distortion levels, even with the time-varying behavior of this type of source. Low level of parameterization is required, since the harmonics present in the signal are identified and the interharmonics are detected, performing component extraction adapted to the window content. The method did not present the lowest RRMSE for the time invariant simulated signals, however, it kept the error below 3, 5.10−3. The proposed method presented the lowest RRMSE for time-varying signals compared to the other methods analyzed, among them, with a value lower than 5.10−3 for a 60 % sag. Although the SOBI algorithm was used for the purpose of separating the independent components, the method allows the possibility of using another blind separation algorithm, giving greater implementation flexibility. Thus, the method proved to be capable of being used for a wide variety of signals.