Support Vector Regression Model Predictive Control Based on LM Algorithm and BA

In the system of nonlinear model predictive controller (NMPC) based on support vector regression (SVR), it is difficult for the controller to solve the multi-step control variable according to the predictive model. This paper proposes a one-step or multi-step optimal control variable by rolling the objective function by the Levenberg-Marquardt (LM) algorithm. Meanwhile, the Bat Algorithm (BA) is introduced to optimize the three parameters of the SVR model in order to ensure the accuracy and generalization of the SVR model. From the result of simulation experiments, SVR has better generalization than RBF neural network; the controller based on SVR-NMPC has better adaptability and robustness in tracking performance as well as the capacity of anti-interference and anti-noise, compared with PID controller.