【摘要】 最小二乘支持向量机(LS-SVM)方法克服了经典二次规划方法求解支持向量机的维数灾问题,适合于大样本的学习。提出一种新的基于LS-SVM模型的预测控制结构,对一典型非线性系统-连续搅拌槽反应器(CSTR)的仿真表明,该控制方案表现出优良的控制品质并能适应被控对象参数的变化,具有较强的鲁棒性和自适应能力。更多还原

【Abstract】 Least Squares Support Vector Machine(LS-SVM) is one of the SVM method which can overcome the dimension disaster of the classic quadratic program method to train the support vector machine,it is fit for the training of large scale data.A novel prediction control structure based on LS-SVM identifier is proposed.A simulation of classic nonlinear process,Continue Stirred Tank Reactor(CSTR) is taken to demonstrate that this method can adapt the changes of the object and has perfect control performance,strong robustness and self-adaptive ability,which is significantly superior compared to the Neural Networks based prediction control and traditional PID control.更多还原