Correct prediction of electricity consumption is important for planning its production in the short term, but also in the long term due to the construction of new power plants and mining planning. Accurate prediction is a challenging task because the consumption changes both in the day and during the whole year. The paper describes a method based only on input data for consumption. No additional influences were included such as temperature, wind, GDP (Gross Domestic Product). Five machine learning algorithms were used to create a predictive model. The best results were achieved with a local polynomial regression algorithm. Daily prediction error was 5.77%, weekly 3.49% and monthly 2.41%.