Abstract：To solve the defects of conventional inverse model identification methods for twomotor variable frequency speedregulating system, a multimodel least squares support vector machines (LSSVM) modeling method was proposed. Based on multimodel, the LSSVM was used to fit the nonlinear mapping of local inverse models. The estimation of every LSSVM local model was fused by computing the weighted sum of local models to build the initial offline inverse model of the system. According to system deviation, the modified recursive leastsquares (RLS) algorithm was adopted to adjust the weights adaptively. The initial inverse model was corrected online to improve identification accuracy and convergence rate. The twomotor variable frequency speedregulating system and the corresponding invertibility were analyzed, and the improved RLS algorithm and LSSVM algorithm were described. The steps of realizing multiplemodel LSSVM identification scheme was provided to achieve the simulation of identification effect. The simulation results show that the proposed method is suitable for the inverse model identification of twomotor variable frequency speedregulating system.
2013, Vol. 34 
