Transient Stability Assessment of High Proportional Renewable Energy Power System Based on Improved Bi-LSTM under Data Sample Imbalance Distribution

The imbalance between the quantity of stable and unstable measured data in power system may lead to prioritizing stable samples for the data-driven transient stability assessment model training. Thus this will easily cause the missed judgment of unstable samples and cause serious consequences. To conquer this shortage, an improved Bi-directional Long-Short-Term Memory (Bi-LSTM) is proposed in this paper, which is adopted to estimate the transient stability of the highly proportional renewable energy power system. The proposed evaluation model guides the training process by introducing a class-balanced focal loss function, which uses the inverse of the effective sample number as the loss weight. Therefore, this loss function has suitable weights among different categories, making a smaller quantity of sample categories with higher loss weights to solve the tendency of model training and ensure the stability and reliability of the model. In addition, the performance of the proposed model is also analyzed under different classifier output thresholds. Finally, simulations are carried out on the New England 10-generator 39-bus system connecting renewable energy generation. The results verify that the model proposed in the paper has superior performance than other machine learning algorithms. It also proves that the selection of the classifier output threshold will affect the accuracy of the transient stability assessment model.