To effectively use the empirical knowledge to compensate for incomplete training data coverage, a fuzzy
modeling method that incorporates empirical knowledge in the form of TSK(Takagi-Sugeno-Kang) fuzzy rules is proposed.
In the structure identification process, a fuzzy rule fusion method is proposed to determine the initial fuzzy rules. In the
parameter identification process, the original objective function of the gradient descent method is improved and the evaluating
parameter of the accuracy of empirical knowledge is introduced to trade off the influence of sample data and empirical
knowledge. The numerical simulation and engineering case studies show that the proposed method can offer more reliable
and accurate forecasting values.