Parallel Multiscale CNN-LSTM-Based Multimodal Data Stream Transmission Demand Prediction

Power equipment fault detection is critical for reliable grid operation. Traditional methods rely on a single data modality with inadequate feature representation. The growing popularity of the Internet of Things (IoT) promotes multimodal data fusion, in which image and voiceprint data play a crucial role in monitoring. Therefore, this paper proposes a parallel multiscale convolutional neural network-long short-term memory (CNN-LSTM)-based multimodal data stream transmission demand prediction. Firstly, we construct a multimodal data model for IoT-enabled power equipment monitoring. Then, a parallel multiscale CNN-LSTM-based multimodal data stream transmission demand prediction algorithm is proposed. The algorithm first extracts image and voiceprint data features using the parallel multiscale CNN module on the edge server. It then predicts the future data stream demand through time-series modeling with the LSTM module. Simulation results show that the proposed algorithm improves prediction accuracy by 2.66% and 15.01% compared to Baseline1 and Baseline2, respectively. This verifies the effectiveness of the proposed method and provides technical support for the intelligent operation and maintenance of power systems.