摘要
赖氨酸发酵过程是一个复杂的非线性强耦合动态过程。某些发酵过程关键生物参数(如菌体浓度、基质浓度、产物浓度)难以实时在线检测。采用软测量技术可解决这一难题。建立了用于生物参数状态预估的tPSO-BPNN软测量模型。针对BPNN软测量模型易陷入局部极小值,进化后期收敛速度慢以及全局搜索能力弱等缺陷,tPSO-BPNN软测量模型采用带极值扰动粒子群(tPSO)算法优化BP神经网络权值和阈值。仿真结果表明,tPSO-BPNN软测量模型的性能优于BPNN软测量模型,能够准确预估赖氨酸发酵过程中的关键参数,具有较高的精度和良好的应用前景。
Lysine fermentation process is a complex nonlinear dynamic strong coupling process. It is very difficult to measure the primary biological parameters on line. However, soft sensing can solve the above problem. Aiming at the disadvantages in BPNN soft sensor, such as falling into local minimum easily, slow convergence speed and weak global search capability, a soft sensor model based on tPSO-BPNN is proposed and used for estimating the biological parameters, in which tPSO algorithm is applied to optimize weights and thresholds of BP neural network. Simulation results show that the proposed model could measure the key parameters on line during the course of lysine fermentation. And the model based on tPSO-BPNN has higher precision and better performance than the model based on BPNN.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2010年第10期2317-2321,共5页
Chinese Journal of Scientific Instrument
基金
国家"863"计划(2007AA04Z179)
高等学校博士学科点专项基金(20070299010)资助项目
关键词
带极值扰动粒子群优化
赖氨酸
生化参数
软测量
extremum disturbed particle swarm optimization （tPSO）
lysine
biochemical parameter
soft sensor
