Voltage sag source classification identification based on AST-CEDT

Abstract

To satisfy the requirement of the accuracy and efficiency of voltage sag features detection and disturbance source identification introduced by the power quality management, an approach for identifying voltage sag sources based on AST (adaptive S-transform) and CEDT (chaotic ensemble decision tree) is proposed. Firstly, the multi-resolution analysis of single and compound voltage sag disturbance signals is carried out by using AST which can adjust the window width factor adaptively. Then, eight feature quantities of disturbance are extracted from the AST modulus matrix according to the characteristic differences of the above disturbances. Finally, a chaotic ensemble decision tree is constructed by combining the ergodicity advantage of Tent chaotic search and the “collective intelligence” of ensemble decision tree to accomplish the effective identification of voltage sag disturbance sources. The simulation results manifest that the proposed method can accurately identify the voltage sag disturbance source signals under different noise environments, and its classification accuracy is higher than that of traditional decision trees, weighted nearest neighbors method, support vector machines and probabilistic neural networks. The new method has good noise resistance and robustness.