A Data Analysis Method for Abnormal Operation of Intelligent IoT System Based on Federated Learning and Deep Learning

Data friendly interaction is a key link in ensuring the smooth and reliable operation of the Power Internet of Things (pIoT), and accurate and efficient identification and analysis of abnormal network data is essential. In response to this, this article proposes an anomaly data identification method that combines federated learning and deep learning. This article uses the Long Short Term Memory Network (BiLSTM) to perform forward and backward learning training on the data features of the pIoT network, deeply and accurately mining and learning the information features of the data. Furthermore, horizontal federated learning is introduced to achieve adaptive parameter adjustment and precise model fitting of the network model, achieving accurate identification and analysis of data states. The simulation experiment used the CIC-ToN-IoT dataset to simulate the complex situation of the intelligent IoT system in the power grid. The results showed that the proposed method had identification indicators RMSE and MAE of over 91.18% and 91.34% for complex datasets, respectively, and had excellent ability to identify and analyze abnormal data.

I. Introduction

The emergence of the Internet of Things effectively integrates the data and information flow in power systems, achieving effective and precise control of systems and equipment [1]. The current scale of the power system is expanding, terminal devices are becoming massive, and business data is becoming complex, posing significant challenges to the power Internet of Things.