样本不平衡情况下的电力系统暂态稳定集成评估方法

In order to quickly and accurately evaluate the stability of the power system after a transient fault occurs in the power system, and to solve the bias problem of the model caused by sample imbalance, an integrated transient stability assessment method for power systems based on the improved loss function is proposed. Firstly, based on the short-term measurement data after the fault clearing, a new integrated model that combines one-dimensional, two-dimensional single-channel and two-dimensional multi-channel convolutional neural networks is designed to realize the end-to-end abstract feature extraction and transient stability classification. Secondly, the loss function in the model training process is improved to enhance the fitting degree of unstable samples for increasing the weights of the misclassification samples. Thus, the global accuracy is improved, and the missing alarm rate of unstable samples is reduced. Moreover, the influence of the output threshold of the integrated model on the recall rate of instable samples is analyzed in this paper. Finally, the simulation results of IEEE 39-bus system and IEEE 145-bus system verify the effectiveness of the proposed algorithm.