Dynamic Time-Series Clustering for Improving Appliances Forecasting in Smart Homes

Energy consumption forecasting for smart home appliances is a crucial area of research due to its potential to optimize energy consumption and reduce costs. However, accurate energy forecasting remains challenging due to the evolving nature of energy consumption patterns. This paper addresses these challenges by proposing Dynamic Time-Series clustering techniques to enhance accuracy in energy and load forecasting for smart home appliances. The focus of this study is on the correlation between the energy consumption of different household appliances. Energy consumption data from these appliances is recorded as time-series data, offering valuable insights for analysis. Our research focuses on clustering time-series data from multiple household appliances to enhance the performance of forecasting algorithms. Specifically, we aim to improve forecasting accuracy by using clustering information. This involves organizing time-series data from various appliances into clusters, which are then used as input for the forecasting algorithm. To achieve this, we employ the k-shape algorithm for time-series clustering on household appliances data. This algorithm effectively extracts time series shapes and computes pairwise cross-correlation for clustering purposes. Moreover, our proposed Dynamic time-series clustering approach incorporates time of the day information, enabling the identification of patterns and variations in appliance behavior across various periods.