基于混合粒子群优化的异类多模型软测量方法研究与应用

基于异类组合预测模型可提高模型的预测精度及鲁棒性的思想,提出一种基于混合粒子群优化的异类多模型非线性组合软测量建模的新方法。即先分别用混合粒子群优化的径向基函数神经网络、最小二乘支持向量机及部分最小二乘算法对训练集训练得出子模型,然后将具有性能互补性的三个子模型的输出作为反向传播网络的输入得到最后结果。用混合粒子群优化的方法来选取径向基函数神经网络和最小二乘支持向量机的模型参数,该方法克服了常用的交叉验证法耗时与盲目性问题。三层反向传播网络具有无限逼近特性,使得整个组合预测模型具有更好的泛化能力和预报精度。将其应用于汽油调合系统中研究法辛烷值的预测,仿真结果表明,该方法是可行且有效的。

Research on the Method of Multi-modeling Soft Sensor Based on a Hybrid Particle Swarm Optimization Algorithm

Inspired by the idea of combining models to improve prediction accuracy and robustness,a new method based on hybrid particle swarm optimization algorithm for heterogeneous multi-model nonlinear soft sensing modeling was proposed.A whole training data set was trained by particle-swarm-optimizing radial base function networks(RBFN),least square support vector machine(LS-SVM)and partial least square algorithm(PLS).The output of three submodels with the performance of complementary was used as input of back pro...