Abstract:

For predictive model mismatch and difficulty in solving nonlinear optimization function of nonlinear system model predictive control, an online support vector machine (OSVM)
modeling and genetic algorithm (GA) rolling optimization based model predictive control is proposed. This algorithm modeled nonlinear object by OSVM, which is an iterative support vector regression learning algorithm and can be used for online training, hence predictive model parameters could be adjusted online through online learning; furthermore, nonlinear optimization function is solved by GA optimization, rolling optimization is realized in system. The nonlinear case simulation results showed system adaptability was improved by the present algorithm.