REAL-TIME SMART MCB BASED ON NODE-RED FOR PROTECTION AND ENERGY PREDICTION USING LSTM

Abstract

A household electrical protection system based on the Raspberry Pi 3B was developed to prevent damage to electrical appliances caused by overcurrent and undervoltage faults. Conventional Miniature Circuit Breakers (MCBs) can protect against overcurrent faults but cannot protect against undervoltage, cannot be remotely controlled, and require manual resetting after tripping. Therefore, this study proposes a smart MCB system that provides overcurrent and undervoltage protection, real-time monitoring, Telegram Bot notifications, and one-day-ahead electrical energy consumption prediction using the Long Short-Term Memory (LSTM) method. The proposed system uses a Raspberry Pi 3B as the main controller, a PZEM-004T sensor to measure electrical parameters, a 5 V relay module to drive an LY2N relay, and the LY2N relay to disconnect or reconnect the load. The PZEM-004T sensor achieved average measurement errors of 0.60% for voltage, 7.29% for current, and 0.87% for active power. The overcurrent protection test showed that the system followed an inverse-time characteristic, where the tripping time became shorter as the multiplier increased. For undervoltage protection, the average tripping times were 2.00 s at 180 V and 2.86 s at 198 V. The LSTM model was trained using 13,041 monitoring data samples and achieved an MAE of 0.000572 kWh, an RMSE of 0.001408 kWh, and an R² score of 67.18%. The average prediction error was 7.58%, with a prediction accuracy of 92.42%. The proposed system successfully performed overcurrent and undervoltage protection, real-time monitoring, and one-day-ahead electrical energy consumption prediction.