A Data-Driven Edge Computing Solution for Real-Time Monitoring of Industrial Load Conditions

With the help of advanced information technology, the smart grid will transition to a digitized paradigm. Advanced data analytics for energy system management and optimization can serve as a key technology pillar for improving energy efficiency and reducing carbon emissions. In this paper, a data-driven condition monitoring solution running on an edge computing device is designed and implemented. Firstly, the collected heterogeneous data are transformed into features, then a clustering learning is used to learn the operating states, and finally an anomaly score metric is used to assess the operating behaviors that deviate from the normal condition. The whole process is done in a single board computer to verify the method, and the data comes from a distributed energy project in southern China. Therefore, the solution proposed in this paper can be used for real-time monitoring of industrial loads and alleviate the dependence on communication networks and aggregation management platforms.