Development of Smart Home System Based on Artificial Intelligence with Variable Learning Rate to Manage Household Energy Consumption

The smart home technology currently being built cannot fully support the government's energy conservation policy. In comparison, the controls for intelligent home configurations are only manual and not fully automatic. In this research, An artificial intelligence-controlled smart home device was designed to manage monthly bill energy consumption in this research. ESP32 was used as an IoT system to detect human presence and measure electrical energy consumed. The data was stored on a Raspberry Pi online server. An Android application can track and manage this device. This application had been reviewed using the Black Box method; the results were 100% smooth. Artificial Neural Network (ANN) with Back Propagation (BP) method added with Variable Learning Rate (VLR) implemented using python language, with four inputs, two layers, and four outputs, each with four neurons. ANN's input variables are light intensity, room temperature, room usage, time length, and monthly cost goal. This ANN-BP-VLR's output is the period of electrical equipment use, in this prototype, air conditioner, TV, refrigerator, and light usage time. The IoT systems for Manage Household energy consumption operated correctly, and the instructions for the optimal use of electrical equipment with an error rate of 17.21%.