MLP-Based Power IoT Terminal Protocols Feature Selection and Protocols Classification Method Research

With the rapid advancement of mobile internet and AI, smart power IoT has gained traction for its flexibility and real-time performance. However, the proliferation of connected terminals brings management complexities and security risks due to protocol heterogeneity and unknown protocols. To address this, this paper constructs an enterprise terminal protocol dataset from manufacturer documents, preprocesses data via numerical transformation and normalization, and uses PCA for feature selection—focusing on message function codes and register addresses. An MLP-based protocols classification method is built and simulations show that MLP achieves 96.2% accuracy with feature selection, which means that this method can play an important role in local identification and registration of existing terminals, laying a foundation for the identification, protocol analysis, presentation, and device management.