A novel Hypertuned Prophet based power saving approach for IoT enabled smart homes

Recent advances in energy conversion, information technology, the internet, new types of web, and communication technologies have enabled the interconnection of all physical objects, including sensors and actuators. Web-enabled smart objects have paved the way for smart homes by enabling innovative services. In this work, the idea of using time series analysis based on machine learning and forecasting considering the weather conditions is discussed, to enhance automation and improve intelligence. The proposed Prophet model is used to predict the future net energy consumption and generation for improving energy efficiency and enabling power backup. Energy efficiency is the need of the hour, wherein major utilization of energy is in the residential sector, it is gravely important to analyze current generation and consumption and act accordingly for the future predicted results. Moreover, smart homes are dependent on web technologies and telecommunication for the operation of every action, which makes it crucial to have necessary power backup. The Prophet forecasting model, after parameter tuning and logistic growth pattern with additional regressors, gives only 0.27 mean absolute error and 0.13 mean squared error for predicting future energy consumption as compared to other models.