A new method based on artificial neural network, Wavelet Transform and Short Time Fourier Transform for Subsynchronous Resonance detection

Abstract Subsynchronous Resonance (SSR) is a big concern for generator protection in power systems, especially with the wide deployments of series compensated lines. However, detecting SSR in an accurate and fast manner is a very challenging task. Various detection schemes were proposed over the years and they are summarized in this paper. Existing detection signal input options and signal processing techniques are reviewed. Furthermore, this paper proposes a new real time scheme based on Artificial Neural Network (ANN), Wavelet Transform (WT) and Short Time Fourier Transform (STFT) for accurate and comprehensive SSR detection. In the proposed method, signatures are extracted from the electrical and mechanical signals in the first step. Then an algorithm is designed to rank all the signatures and pick up the best performing subset of signatures. ANN is built from signatures contained in the best performing subset. IEEE second benchmark model for SSR study is used to simulate the signals for study in EMTP-type simulation environment. Test results with the generated signals show that the new SSR detection scheme achieves desirable speed, dependability and security for SSR detection. Further research will be done to apply the selected techniques and signatures in microprocessor relays.