Fuzzy inference models for short-term load forecasting with tabu search

In this paper, an optimal fuzzy inference approach to short-term load forecasting in power systems is proposed. The proposed method provides the optimal structure of fuzzy inference that optimizes the number and the location of the fuzzy membership functions so that the model errors are minimized. For simplicity, the simplified fuzzy inference is used to handle the output variable as the crisp number. Short-term load forecasting is one of the most important problems in power system operation and planning. High accuracy of the load forecasting improves system security and saves generation cost. The advantage of the fuzzy inference over ANN is that it allows to analyze the inference process with the fuzzy rules. Since the relationship between input and output variables is clarified, the results are intuitively understandable. However, the open problem is how to construct optimal fuzzy systems with respect to the number and the location of the fuzzy membership functions. This paper presents a new method for constructing the optimal structure of fuzzy inference with tabu search. Also, the parameters at the consequence are optimized with supervised learning. Fuzzy models with different input variables are tested in actual data