Analysis of formalized methods for forecasting the volume of electricity consumption

We consider the construction of basic formalized forecasting methods. The methods are widely applied in the world markets of electric power industry. The models are tested on the actual hourly data of the United energy system of the Wholesale Electricity and Capacity Market of Russia. Each model is tested for adequacy by means of Fisher's F-test, t-statistics, mean error of approximation and determination coefficient. We construct a correlogram in order to prove a cyclicity of the time series of electricity consumption, where periods are 1 week and 1 year. An autoregression of the 7th order, that is, with cyclicity in a week, shows the highest efficiency among the examined models. A multifactorial linear regression, taking into account 3 external factors (market rate per day forward; average daily ambient temperature; qualitative factor of working and non-working days), has the least efficiency. The proposed methods are recommended for the operations of subjects of electric power industry to forecast the main parameters of energy market in order to reduce the penalties by improving the forecast accuracy.