Predicting electricity consumption
The article is concerned with an approach towards a short-run prediction of electricity consumption based on macroeconomic indicators and natural conditions. A country’s electricity consumption as a whole is difficult to seize, therefore the researcher has to break it into two groups: households and businesses. Each of these macroeconomic subjects has different determinants, as we cannot expect e.g. the households be very much affected by a growth of the industry production index, not only because the household’s consumption is quite inelastic. On the contrary businesses are about to be easily influenced by economic cycles and new market situations.
The prediction itself is then reached through combination of ordinary least squares regression model and a time series analysis using exponential smoothing. The regression is used for estimation of significant drivers of the electricity consumption, when exponential smoothing helps to extrapolate the determinants used in the model. The final course of the electricity consumption is then a weighted average of the regression results and extrapolation of the electricity consumption time series itself. Practical example of the approach will be shown on the data from the business sector of the Czech Republic.