Harmonic detection methods for active power filters based on discrete cosine transform and dithering

Harmonic detection is a very important factor in control of active power filters, since it generates the reference signal for the current controller. The performance of the whole system will be affected by the suitability of the implemented harmonic identification method. In this work, the equivalences and advantages of the discrete cosine transform with respect to other Fourier based algorithms are demonstrated. It is proved in this paper that both the transient length and the phase error in presence of grid frequency deviations are halved in comparison to DFT. Therefore, it is convenient to provide a frequency adaptation feature to the discrete cosine transform. In this work it is proposed to adjust the window width to the grid fundamental frequency, and to enhance the frequency resolution by means of dithering. The theoretical approach is verified by experimental results in an APF lab prototype.