基于PSO-SVM的关键生物参数软测量方法研究

Study of Soft-sensing Method Based on PSO-SVM for Key Bio Parameters

微生物发酵过程中一些关键生物参数难以实时在线测量,严重影响发酵的优化控制。为解决关键生物参数的测量难题,采用了一种基于PSO-SVM的软测量方法。该方法利用粒子群优化(PSO)算法优化选择支持向量机(SVM)的最佳参数,并建立了基于PSO-SVM的软测量模型。利用赖氨酸发酵的数据对模型进行仿真验证,结果表明该模型具有很好的学习精度和泛化能力。另外在建模耗时上,PSO-SVM算法所用时间远少于标准SVM算法所用时间。

Some key biological parameters in the process of microorganism fermentation are difficult to measure online and real-time,which greatly affects the optimization of control system in fermentation.In order to solve the problem of measurement,this paper used a soft-sensing method based on PSO-SVM.The optimal parameters of support vector machines（SVM）were optimized by the algorithm of particle swarm optimization（PSO）.Meanwhile,the soft-sensing model based on PSO-SVM was established.Then the soft-sensing model was verified by simulation according to Lysine fermentation data,the simulation results show that the model has good learning precision and generalization ability.The modeling time of PSO-SVM algorithm is far less than that of SVM.