Variant regression modeling of electricity production in the Russian Federation

Abstract

Objective. The aim of the study is to build a linear regression model of electricity generation in the Russian Federation depending on resource indicators, which include: the volume of coal and gas production, the production of fuel oil. Statistical data for 2005 - 2020 were used as the information base of the study.Method. Estimation of unknown parameters of the linear model is carried out using three methods - least squares, modules and anti-robust estimation. They behave differently with respect to outliers in the data. The second of them does not react to outliers at all, completely ignoring them, and the third, on the contrary, strongly gravitates towards them, therefore, these methods are a kind of antagonists in relation to each other.Result. Three alternative models of a linear regression model of electricity production with high accuracy are obtained. The value of the parametric stability index of the data sample, based on the properties of the parameter estimation methods, is calculated. Observations are identified that correspond to the maximum and minimum extent to the linear model on the analyzed sample. The values of the contributions of the factors to the right parts of the models are calculated.Conclusion. Three versions of the model built by different methods can be successfully used to solve problems related to forecasting the production of electricity in the country. At the same time, the variant constructed by the least squares method is a kind of compromise.

Keywords

• electricity generation
• linear regression model
• least squares
• moduli
• antirobust estimation
• parametric homogeneity index
• factor contributions