Forecasting smart home electricity consumption using VMD-Bi-GRU

Due to its important role in smart grids, power system management, and smart buildings, energy consumption forecasting has gained a lot of interest in recent years, further achieving energy efficiency objectives, decreasing CO2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$_2$$\end{document} emissions, and reducing energy bill. Because of the nonlinear and non-smooth characteristics of residential building electricity consumption time series data, developing an accurate energy consumption model is a crucial task. To solve this constraint, this research proposes a short-term, hybrid model that combines variational mode decomposition and Bi-GRU with the aim to predict household energy consumption forecasting of the next 24 hours with a time granularity of 15 minutes. The VMD algorithm in this model decomposes the power consumption time series into distinct signals called IMFs, and the Bi-GRU is used to predict each IMF separately. To produce the final prediction output, the prediction results of each model are summed and rebuilt. The conclusive findings indicate that the forecasting model based on VMD-BI-GRU demonstrates exceptional performance, with a mean squared error of 0.0038 KW, a mean absolute error of 0.046 KW, a mean absolute percentage error of 0.11%, and a notably high R2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^2$$\end{document} score of 0.98. These results collectively signify its precision as a prediction model.