Intelligent Residual Current Monitor Application in Hospital With Fiber Optics Using Wavelength Division Multiplexing

In this study, we propose the implementation of integrated leakage current detection in residual current monitor (RCM) to control safety protection in microgrid-based hospitals. Using memory-based features to magnify statistical significance, we tried different sampling periods to build a classification model from a sample of RCD induction voltage dataset using support vector regression (SVR). The model was then tested by individually comparing sampling lengths from 2 to 16. The proposed method can also achieve fast and valid detection according to the signal pattern within 5.45ms for H-L sequence with H;high and L;low. By adding the time series method using long short-term memory (LSTM), the detection speed could be further confirmed using H-L-H sequence with root mean square error (RMSE) around 1.% The use of Fiber Optics (FO) can also ensure coordinated control of the microgrid. The detection system can be carried out without delay by separating the machine learning at the server and the digital residual current device (RCD). The digital RCD was synthesized using a Xilinx field-programmable gate array (FPGA) based on XC6SLX16-2FTG256C. The proposed method uses only 184 Registers and 1,656 lookup tables (LUTs). The results present the classification using a designed digital RCD to detect possible leakage current condition (LCC), data delivery to the server via FO using wavelength division multiplexer (WDM), and LSTM acceleration to support the microgrid protection system. This new approach will help develop load fault protection via FO to maximize communication speed, and improve patient safety by localizing the potential hazards.