Complex Fault Source Identification Method for High-Voltage Trip-Offs of Wind Farms Based on SU-MRMR and PSO-SVM

Abstract

Large-scale high-voltage trip-offs (HVTOs) of wind farms are serious incidents afflicting power systems that can lead to voltage instability, power deficiency, and frequency fluctuation. In order to reduce the influence of HVTOs, it is necessary to efficiently identify the fault source after an HVTO at a wind farm. A fault source identification method for wind farm HVTOs is proposed in this work. First, the fault tree analysis (FTA) method is used to summarize the causal and logical relationships among the different factors that lead to HVTOs at wind farms. An index system is constructed according to simulations of wind farm HVTOs under multiple scenarios. Second, a set of feature indices are selected from the original index system as key criteria for wind farm HVTOs based on the symmetrical uncertainty of mutual information and the maximum relevance and minimum redundancy (SU-MRMR) method. Finally, the particle swarm optimization (PSO) method is effectively utilized in the parameter optimization of support vector machine (SVM) method, and the optimized SVM with good performance is further employed for fault source identification based on the feature indices. Both single fault source and compound fault source are identified in an actual power system, and the proposed method is verified as a reliable solution for complex fault source identification for wind farm HVTOs based on the statistical identification accuracy.