Online single-factor measured active nodal load forecasting in an electric power system

Two techniques for online nodal load (NL) forecasting using preliminary classification of training set data are proposed. In the first one, a pattern recognition method, the rate evaluation algorithm (REM), is applied to measured load values of the previous day to classify load diagram that is being forecasted. Diagrams from resulting class are used to calculate load predictions. In the second technique, measured load values of a diagram from training set, which is the closest to the one being predicted, are used as estimates of predicted load values. Online NL forecasting using the mentioned above methods has been conducted. The corresponding mean square errors are given.