Forecasting energy consumption time series using machine learning techniques based on usage patterns of residential householders

• Published in: Energy (Oxford) 2018-12, Vol.165, p.709-726
• Main Authors: Chou, Jui-Sheng, Tran, Duc-Son
• Format: Article
• Language: English
• Subjects: Artificial intelligence; Buildings; Data mining; Energy consumption; Energy costs; Energy efficiency; Energy management; Environmental impact; Forecasting; Learning algorithms; Machine learning; Model accuracy; Optimization; Residential buildings; Residential development; Residential energy; Smart grid; Social change; Statistical analysis; Sustainable development; Time series; Time-series forecasting; Urbanization
• ISSN: 0360-5442; 1873-6785
• DOI: 10.1016/j.energy.2018.09.144

Energy consumption in buildings is increasing because of social development and urbanization. Forecasting the energy consumption in buildings is essential for improving energy efficiency and sustainable development, and thereby reducing energy costs and environmental impact. This investigation presents a comprehensive review of machine learning (ML) techniques for forecasting energy consumption time series using actual data. Real-time data were collected from a smart grid that was installed in an experimental building and used to evaluate the efficacy and effectiveness of statistical and ML techniques. Well-known artificial intelligence techniques were used to analyze energy consumption in single and ensemble scenarios. An in-depth review and analysis of the ‘hybrid model’ that combines forecasting and optimization techniques is presented. The comprehensive comparison demonstrates that the hybrid model is more accurate than the single and ensemble models. Both the accuracy of prediction and the suitability for use of these models are considered to support users in planning energy management. •A comprehensive analysis of ML for forecasting energy consumption in buildings.•The advantages and disadvantages of presented ML method are analyzed.•Prediction accuracy and suitability for use of ML are considered in this investigation.•The hybrid model is more accurate than the single and ensemble models.•Concluding remarks support users in planning energy management.