I-CNN-LSTM: An Improved CNN-LSTM for Transient Stability Analysis of More Electric Aircraft Power Systems

Abstract

High-power nonlinear load characteristics are one of the typical characteristics of multi-electric aircraft power systems. The study provides an improved CNN-LSTM stability analysis method for solving the stability problem of the aircraft power system caused by high-power nonlinear load switching. To address the issue of sample imbalance, this approach creatively incorporates the cost factor into the CNN loss function. In order to handle the issue of computational complexity, the projection layer is added to the LSTM, and a methodology known as CNN-LSTMP is proposed. This algorithm solves the problems of low computational efficiency and huge computational volume. The time series data utilized by the experiment are created by simulating the transient switching process. The data are then labeled, normalized, and model training is carried out. A deep learning algorithm that satisfies the prediction requirements can be created by applying this method to the established simulation model of a multi-electric aircraft power system for stability analysis. According to the results of the experiments, this method’s transient stability analysis accuracy is 93.32%, which has a positive impact on transient analysis and may satisfy application requirements.