Data mining for short-term load forecasting

Short-term load forecasting plays a key role in power system operation and planning. This paper presents a method for data mining for short-term load forecasting in power systems. This paper makes use of a data mining method to clarify the nonlinear relationship between input and output variables in short-term load forecasting. Data mining discovers useful knowledge and rules in large data bases. Data mining is more attractive because of difficulty in understanding large data bases. The obtained model structure explains the importance of input variables. It may be classified into the classification and the regression trees. This paper handles the regression tree since load forecasting corresponds to the quantitative problem. This paper presents three strategies: hybrid model of CART and multi-layer perceptron (MLP); optimal structure with Tabu search; and fuzzy data mining.