Integration of Deep Learning into Smart Distribution Panelboard at Household level with IoT Connectivity

Power management systems are becoming inevitable for enhancing efficiency, reliability, and sustainability in household electricity consumption. Unlike industry which has a robust intelligent power management system for effective control. The proposed system introduces an integration of IoT with cutting edge deep learning modality using the facilitation of microcontroller as entry-exit core point edge device to receive data from the sensor modules, conveying data to a visualization platform through Wi-Fi (MQTT protocol). In order to address the concern of power wastage due to unawareness of power consumption it needs to enable the visual portal which displays the power, current and voltage consumption of the household by deploying it as an application that could be scaled and installed into local mobile phones for real time monitoring. The sensor modules are embedded onto a distribution panel board prototype sensing the sample load in various distribution nodes and circuit interpreters to shut down during abnormality. In the second phase, these data are then transmitted to a central control unit for decision-making using advanced deep learning algorithms. The comparison of various deep learning algorithms like light gradient boosting machine (LGBM), neural network (NN) and long short-term memory (LSTM) tabulating their accuracy in a more deterministic aspect using the mean absolute percentage error (MAPE) as evaluation parameter attempting to be more data centric artificial intelligence. The prediction comes under the temporal categorization of load forecasting attempting to predict on a short term basis for effective system management.