Communication technology that utilizes power lines can substantially elevate the intelligent management capabilities of smart grid infrastructure and bolster the operational efficiency of the grid. A Two-Way Automatic Communication System (TWACS) enables relay-free signal transmission across transformers, thereby enhancing system robustness. However, the accuracy of modulated signal identification in traditional TWACS implementations is often inadequate. Consequently, this paper introduces a novel signal recognition approach that employs a Bidirectional Long Short-Term Memory (Bi-LSTM) neural network optimized by the Dandelion Optimization Algorithm (DO) and augmented with an attention mechanism. Initially, the model employs a convolutional neural network (CNN) to adaptively extract significant features from TWACS signals. Subsequently, it utilizes the DO method to perform hyper-parameter optimization for the Bi-LSTM. With the optimized parameters, the model constructs the neural network and integrates an attention mechanism to assign varying weights to the input features. This integration enhances the model’s capacity to recognize signals more effectively. The proposed method is capable of recognizing modulated signals even under conditions of low signal-to-noise ratios. The resultant model, designed for signal identification, satisfies the fundamental requirements for real-time processing. When applied to analyze transmission signals from distribution areas with new energy access, the proposed algorithm achieves a recognition rate of 92.32%. This high rate indicates that the method can efficiently and accurately identify TWACS modulated signals, thereby offering robust support for the integration of power line communication technology into smart grid applications.
Hua et al. (2026) studied this question.