Multiple Model Controller Based On M2SPSA

This paper provides a new method to control the model-free industrial process plants, which is based on the Modified Adaptive Simultaneous Perturbation Stochastic Approximation(MASPSA/M2SPSA) Algorithm and the Multiple Model Approach. This method has good prospects in the control of the model-free industrial objects because of the M2SPSA's simple algorithm structure, capability in dealing with noisy measurements and effective manipulation in the control of model. And this is also because of the Multiple Model Approach's enhancement of transient response and its effective control of large undetermined change in system. It is more effective to combine the M2SPSA algorithm and the MMA. In practice, we can use the differences of inputs of object and the controllers' outputs to create a target function. The target can select the optimal controller of the system in the multiple controllers. The adaptive controller which is based on the M2SPSA can use the differences of inputs of object and the controllers' outputs to obtain another target. M2SPSA Algorithm is used in the minimizing of this target. And we use a wise method to select the system's reference inputs, which is called Virtual Model Reference Adaptive Control.The intervention of Multiple Model Approach improve the whole system's rapidness, robustness and stability. The parameters of the multiple fixed model controller can re-initialize the adaptive controller which is based on the M2SPSA algorithm. Simulation experiments for the general industrial process demonstrate the validity and practicability of this method.