This paper presents using of artificial neural network for short-term and medium-term forecasting of electric power consumption by energy provider. As a program for modeling an artificial neural network we chose MATLAB with its Neural Network Toolbox. The article reviews two training algorithms. First was the Levenberg-Marquardt algorithm and the second was Bayesian regularization algorithm. Comparing this algorithms was founded that the second one Bayesian regularization algorithm is more effective for making short-term and medium-term forecasts of electric power consumption. Also it is found, that artificial neural network can be used for one-day forecast management accurate within 2.5% from actual consumption. Comparing the forecasting data with actual data of electric power consumption allowed to the conclusion that artificial neural network which was chosen can be used on practice for forecasting of electric power consumption of energy provider for its effective work on the wholesale electricity market.
Keywords: energy provider, forecasting, plan-by the hour forecast, data base of energy consumption, wholesale electricity market, dynamic series, artificial neural network, Levenberg-Marquardt algorithm, Bayesian regularization algorithm, neural network architectur