A novel approach to estimate building electric power consumption based on machine learning method: toward net-zero energy, low carbon and smart buildings

Abstract

Abstract The modern era has witnessed a surge in energy consumption and its dependence on fossil fuels, which are harmful to the environment, prompting researchers to examine techniques for regulating energy usage in buildings, specifically with regard to residential electricity consumption. The pursuit of net-zero energy consumption and low carbon emission buildings is a significant undertaking that nations across the globe are actively endeavoring to accomplish. In order to accomplish this goal, the structure in question must efficiently oversee its overall energy usage while concurrently capitalizing on sustainable energy sources. The precise estimation of future electricity usage in buildings is an essential element in the process of energy efficiency planning and optimization. The present study introduces a soft computing methodology and data decomposition as approaches for evaluating the energy usage of residential structures. An innovative machine learning approach is introduced for the purpose of estimating the initial cost required to construct a green structure that consumes no net energy. By utilizing wavelet decomposition, it is possible to determine how to transform the structure into one that is intelligent and energy efficient. Following wavelet parallel converter analysis, the data were processed with an estimator model based on an ideal neural network. The results indicate that the mean estimation errors for recurrent neural network, Autoregressive fractionally integrated moving average (ARFIMA), and gene expression programming (GEP) were reduced by 72%, 65%, and 77%, respectively, using this method. Conversely, when the proposed methodology is applied to the smart management of building energy consumption, the examined structures experience an average reduction of 8% in energy consumption. Moreover, the outcomes of CO2 gas emissions demonstrate that the suggested model possesses the capability to accurately forecast CO2 emissions. The study highlights the necessity of employing innovative techniques such as machine learning to decrease building energy usage and CO2 emissions. The discovery of these results can assist policymakers and stakeholders in the energy sector in advancing the adoption of smart building technologies.