A Power IoT Terminal Protocols Identification Method via Decision Trees

With the increasing number of smart power IoT terminal devices, the management difficulty of these terminals, such as smart meters, inverters, sensors for wind speed, temperature and humidity, is growing. The intelligent control for the current power Internet of Things can only cover edge devices, while the identification capability of the terminals is weak. The workload for grid operators and maintainers is rising daily. To address issues such as the diversity of protocols from different device manufacturers, low automation levels, high labor costs, and management inefficiencies, this paper analyzes the traffic protocol of various power IoT terminals and extracts the application layer payload data. Based on the preprocessed data and according to the characteristics of message loads application data in different protocol types, the effective features closely related to message byte length and inter message distance information are extracted. Furthermore, a method for identifying and classifying power IoT terminal protocols using classical machine learning based on decision trees is proposed, laying a foundation for subsequent protocols parsing, presentation, and devices management. In this paper, the experimental results demonstrate that this method achieves a classification accuracy of over 97% for protocols of devices such as 104, 698, and Modbus.