Electric Load Forecasting On Smart Energy Meter (SEM) Using Linear Regression

Smart Energy Meter (SEM) is a device that can monitor and forecast electricity usage in the future. This SEM is based on the concept of a conventional electricity meter but has been modified with the addition of a microcontroller to acquire and process data such as voltage, current, frequency, and power factor. The data collected by the microcontroller is sent to a WiFi module for transmission to a local database. Once the data reaches the database, it is processed and calculations are performed to obtain active power (P), apparent power (S), reactive power (Q), phase angle, and energy consumption. After collecting the data, linear regression equations are used to forecast future electricity loads. The testing for load forecasting in this SEM involves collecting data from the load panels in the Electric Drive and Power Electronics Laboratory, Department of Electrical AutomationEngineering, Vocational Faculty, Sepuluh Nopember Institute of Technology. Data is collected every minute from 8:30 AM to 3:30 PM for three weeks. The usage data from two weeks is used to forecast electricity load usage in the third week. The results of the testing and analysis, using the Mean Absolute Percentage Error (MAPE), show an average forecasting error of 3.86% for active power (P), 3.77% for apparent power (S), and 11.87% for reactive power (Q) in the third week.