Functional Data Analysis for Household Appliance Energy Consumption Prediction

Over time, the increasing use of home appliances, driven by the industrial revolution, has significantly contributed to overall household energy consumption. The number of appliances and various environmental indicators can also impact energy usage. Therefore, it is important to understand how to optimize energy utilization and improve efficiency. Analyzing energy consumption presents statistical challenges due to the large size, high frequency, complexity, and noise in the data. We investigate the use of Functional Data Analysis (FDA) approaches to address these challenges. Unlike traditional methods that treat each observation as an individual variable contributing to the overall dimension, FDA considers the entire trajectory as a single data curve. In this paper, we provide a step-by-step analysis of functional regression models to quantify the relationship between household energy consumption and several environmental indicators. Our dataset consists of energy consumption recorded in real-time at 10-minute intervals from an observation house. We compare the model performance of our FDA models with linear regression, support vector machine, and random forest. Our empirical results show that functional regression effectively captures the dynamic effects of environmental conditions that vary over time and have the lowest root mean square error and mean absolute percentage error.