论文部分内容阅读
最小二乘支持向量机(LSSVM)软测量建模的回归精度和推广能力取决核函数选取以及超参数的优化,然而多核概念的引入,弱化了核函数的选取,使问题更加倾向于超参数的优化,针对多核LSSVM的超参数优化问题,采用全局搜索能力较为出色的粒子群(PSO)算法,最后并将其应用于污水处理软测量建模,仿真分析证实该方法进一步改善了回归模型的精确度和COD浓度的在线预估能力。
The regression accuracy and popularization ability of least squares support vector machine (LSSVM) soft-sensing modeling depend on the selection of kernel function and the optimization of hyperparameters. However, the introduction of multi-core concept weakens the selection of kernel function and makes the problem more prone to hyperparameter Optimization, aiming at the problem of hyper-parameter optimization of multi-core LSSVM, particle swarm optimization (PSO) algorithm with better global search capability is adopted. Finally, the PSO algorithm is applied to sewage soft sensor modeling. Simulation results show that this method can further improve the precision of regression model Degree and COD concentration of the online ability to predict.