Transmission Line Audible Noise Prediction Based on CNN-BiLSTM-Attention Method

Abstract High-voltage transmission lines’ audible noise parameters are impacted by a variety of multidimensional elements. in order to better the accuracy prediction and effectively utilize time-series properties observed data. In this paper, we propose combined model convolutional neural network (CNN) bidirectional long short-term memory (BiLSTM)-based attention mechanism based on feature filtering for line prediction. Firstly, using real-world data as dataset, factor time series optimally filtered, high correlation vectors extracted CNN. Secondly, fed into BiLSTM training prediction, performance is further improved introducing an at end so that focuses learning more important features. Finally, analysis actual recorded from 500 kV AC Sichuan Province demonstrates CNN-BiLSTM-Attention suggested paper has higher than BiLSTM, CNN-BiLSTM, BiLSTM-Attention models.