四川大学电气工程学院，四川省 成都市 610065
College of Electrical Engineering, Sichuan University, Chengdu 610065, China
This work is supported by National Natural Science Foundation of China (No. 51807126) and Full-time Postdoctoral Research and Development Fund of Sichuan University (No. 2019SCU12003).
It is important to accurately identify voltage sag sources for sag responsibility allocation and mitigation decision-making. This paper proposes a recognition method of voltage sag sources based on the optimized extreme learning machine (ELM). The time-domain features are directly extracted from the voltage sag waveforms, and the time-frequency domain features (include energy entropy and singular entropy) are extracted by S-transform. Then the feature vectors are built based on the time-domain and time-frequency domain features. The feature vectors make up for the shortcomings that the existing methods only use time-frequency transform to extract features and may lose the important information only existing in the time-domain, which will affect the recognition accuracy. The genetic algorithm is used to optimize the input weight and hidden layer bias of the ELM. The optimized ELM model is proposed to solve the problem of the pattern recognition, whose model is complex and time-consuming. The validity of the proposed feature vectors and optimized ELM model are verified by the simulated data and the measured data. Compared with other methods, it is proved that the proposed model is simple and fast in training and classification, and has higher recognition accuracy. It is suitable for edge calculation and can identify the voltage sag sources accurately and fast.
WANG Ying,WANG Huan,ZHANG Shu.Recognition Method of Voltage Sag Source Based on Optimized Extreme Learning Machine[J/OL].Automation of Electric Power Systems,http://doi.org/10.7500/AEPS20190809004.