Sub-Synchronous Oscillation Recognition Based on Extreme Learning Machine

Abstract

In recent years, sub-synchronous oscillation (SSO) occurs frequently in wind power systems in many countries, which seriously affects the safety and stability of power systems. In order to monitor and warn sub-synchronous oscillations when they occur and avoid further escalation of accidents, it is urgent to carry out wide-area monitoring for SSO. Therefore, a sub-synchronous oscillation identification method based on Extreme Learning Machine (ELM) is proposed in this paper. Through the analysis and observation of PMU signal waveform images, this paper found that SSO can be identified from three aspects: the trend of PMU signal waveform, the fluctuation of envelope and the length of stationary subsequence. Thus, the SSO identification problem can be transformed into a Classification problem. In this paper, the classical ELM algorithm is adopted to realize the fast and accurate recognition of SSO. The proposed method is verified in detail by the synthetic signal, and compared with the Support Vector Machines (SVM) algorithm. The results show that the proposed method can effectively identify SSO even in the case of large noise. Therefore, the theoretical results are expected to provide technical support for SSO real-time warning in the future.