PENGEMBANGAN SISTEM PREDIKSI TAGIHAN PENGGUNAAN ENERGI LISTRIK BERBASIS MACHINE LEARNING

• Main Author: Farhan Anshari, Dimas
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/73932
• Institution: Institut Teknologi Bandung

This Final Project is a study that develops a smart energy metering and electricity bill forecasting system. There are four main subsystems built, namely: embedded system circuits and Internet of Things, Android-based applications, backend applications, and machine learning. This report explains the development of a machine learning-based electricity bill forecasting system, which is part of the smart energy metering and electricity bill forecasting system. The basis of this research, in general, is to increase the user's awareness of the rise in electricity costs that must be paid based on its use. This research discusses in-depth the machine learning process using the ARIMA model to predict the electricity costs that the system users will pay in the future. The dataset used was obtained from a data bank (opendatanetwork.com), which is in line with the needs of this subsystem's development. The use of machine learning technology was chosen based on the evaluation metrics generated from testing 3 different models, namely ARIMA (AutoRegressive Integrated Moving Average), SVR (Support Vector Regression), and LR (Linear Regression). The use of data with a seasonal pattern caused the use of ARIMA to be changed to SARIMAX to match the dataset and produce the smallest RMSE (Root Mean Squared Error) score and an accuracy of 92.66%. The SARIMAX (1,1,2)x(1,0,1)[12] model is used in the development of this system based on the lowest AIC value. The final result of this system is a prediction (forecasting) of the electricity bill costs to be paid in the future and is integrated with the API (Application Programming Interface) so it can be accessed by the application backend. The system has been successfully built and integrated with other subsystems so that the electricity bill forecasting system can be run and used.