Application of modern techniques for detecting subsynchronous oscillations in power systems

Detecting Subsynchronous Resonance (SSR) in power systems in a secure, dependable and fast manner is a major challenge for steam turbine generator and doubly fed induction generator protection. Researchers have been applying Artificial Intelligence schemes to solve generally complex, nonlinear or pattern recognition problems. This paper describes an amalgamated scheme which combines Artificial Neural Networks (ANN) and Wavelet Transforms (WT) to provide accurate and comprehensive SSR detection in power systems. A test system based on the IEEE second benchmark model for SSR is built and modified to generate both stable and growing SSR conditions. An approach combining WT and ANN for SSR detection is presented in a detailed manner. Characteristics from generator electrical and mechanical signals readily available to generator protection systems are extracted and different combinations of these characteristics are used to build the detection scheme. The developed SSR detection scheme has been tested with signals generated from an electromagnetic transients simulation, demonstrating desirable security, dependability and speed for SSR detection.