Advanced Edge Computing Framework for Grid Power Quality Monitoring of Industrial Motor Drive Applications

In large-scale industrial machine applications (IMAs) during condition monitoring, all sensor devices can produce raw data up to 15TB of data/week. Transmitting these large high-frequency data sets to the cloud to closely monitor the operational environment and make decisions is not feasible for two reasons: (a) bandwidth and latency issues and (b) higher cost of data transfer and storage. Condition monitoring applications usually extract critical features from raw high frequency sensor signals and discard raw data to mitigate this issue. The computation is carried out on an edge device near to the application hardware and the role of the cloud/remote server is limited to receiving fault types, features, and monitoring. Therefore, this paper proposes an intelligent data capturing methodology with an edge-cloud framework for grid power quality monitoring of the IMAs that only triggers and transmits datasets to the cloud if the raw datasets contain any grid events and/or grid side anomalies. Using dSPACE, grid emulation is carried out virtually. Feature extraction using Short Term Fourier Transform (STFT) is done in the edge device and grid events are detected based on features. The proposed methodology is configured to send raw grid voltage data and features in the Microsoft Azure-based cloud that contain at least one abnormal grid event. Thus, the proposed approach of this paper limits the space requirement in the cloud by 95%, saves data transmission costs, and enables the cloud to run predictive maintenance algorithms.