Forecasting annual electric power consumption using a random parameters model with heterogeneity in means and variances. (15th September 2022)

• Record Type: Journal Article
• Authors: Hamed, Mohammad M.; Ali, Hesham; Abdelal, Qasem
• Is Part Of: Energy. Volume 255(2022)
• Publication Date: 2022-09-15
• Subjects: Electricity demand forecasting -- Random parameters modeling -- Heterogeneity in means and variances -- Linear regression -- Renewable energy -- Econometric models; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2022.124510
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: This paper develops heterogeneity-based econometric models that forecast the yearly consumption of electricity. Unlike previous research, this paper departs from classical econometric regression models and explicitly accounts for unobserved heterogeneity and corresponding interactions. More specifically, three modeling approaches are applied: a random parameter linear regression model (RPLRM), a correlated random parameter linear model (C-RPLRM), and a random parameter linear model with heterogeneity in means and variances (RP-HMV). In addition, a grey model is estimated based on historical consumption data. The estimation results clearly demonstrate that the random parameter methodology is statistically superior to the classical multiple linear regression model. Moreover, based on the test results reported in this paper, as well as the forecasting accuracy measures, the RP-HMV model is shown to be statistically outstanding, with a very low forecasting error (MAPE of 0.04%) compared to the other models, including the grey model. Our estimation results show that two variables, namely the average electricity price and contribution of renewable energy to the national and distribution grids, produced statistically significant random parameters with heterogeneous variances and means. Moreover, the results show that factors including the number of households on the distribution grid, average electric power price, and availability of all-season air conditioners with hot and cold inverter characteristics significantly influenced the electricity consumption. These empirical results reveal that random-parameters models with heterogeneity in their means and variances can provide a detailed analysis of the predictor variables shaping the annual electricity consumption, and they highlight the importance of including unobserved heterogeneity and related interactions in econometric models. Highlights: The yearly electric power consumption is addressed through heterogeneity models. Unobserved heterogeneity in means and variances are considered. A wide range of explanatory variables affects the yearly consumption of electricity. The random parameter methodology is statistically superior to the classical multiple linear regression model. The random parameter model with heterogeneity in means and variances is statistically superior to other heterogeneity models.