School of Electrical Engineering, Beijing Jiaotong University, Beijing 100044, China
This work is supported by National Key R&D Program of China （No. 2018YFB0904500）and National Natural Science Foundation of China （No. 51577009）.
In order to further improve the prediction accuracy of transient stability for power system and give more refined evaluation results, the deep learning is combined with the transient stability of power system. A refined evaluation model of transient stability for power system based on ensemble deep belief network (DBN) with different structures is proposed based on the characteristics of the generator power angle "trajectory cluster" after fault removal. The base classifier DBN of the model can effectively utilize the feature extraction ability of the deep architecture and fully exploit the nonlinear mapping relationship between the input features and the evaluation results of transient stability. Experimental results on the New England 10-machine 39-node system show that this method is not only superior to the shallow learning framework, but also superior to the partial deep learning model. In addition, the ensemble DBN algorithm not only has higher prediction accuracy, but also can effectively evaluate the stability margin and instability level of the system. It shows strong robustness when some information of phasor measurement unit (PMU) is missing and contains noise.
LI Baoqin,WU Junyong,SHAO Meiyang,et al.Refined Transient Stability Evaluation for Power System Based on Ensemble Deep Belief Network[J].Automation of Electric Power Systems,2020,44(6):17-26.DOI:10.7500/AEPS20190528009Copy