A data-driven multiple-model modeling algorithm

Based on the input-output data , a multiple-model modeling method is suggested for the complex nonlinear system. Firstly, fuzzy partition is employed to on-line clustering for the input-output data; and then, the least-squares (LS) algorithm is employed to construct the local model for each clustering, and the parameter of each local model is updated by the new data. The proposed algorithm takes advantage of the TSK model, combines the fuzzy partition and multiple-model modeling, and updates on-line the number of the local model and the parameter of each by the input-output data, so as to realize the on-line modeling for the complex nonlinear system. Simulation result shows the effectiveness of the proposed method.