Load forecasting based on clustering analysis using fuzzy logic

Fuzzy set theory is one of dominant technology in artificial intelligence (AI). Its application in load forecasting is based on periodical similarity of electric load, where the input variables, output variables and rules are the key point. In this paper, a new related coefficient comparison method is introduced to categorize the variables that have influence on electric load into some clustering groups, and their membership functions are set as input variables in the form of natural language using fuzzy set theory. The history data are clustered into output variables through the adaptive neuron-fuzzy inference system (ANFIS) so as to ensure the minimal errors of their member functions. The rules that link the input variables with output variables are set on the basis of practice data and expert knowledge. This method is an effective tool to deal with some special variables such as weekends efficiently and closed to the actual load forecasting in practice.