基于神经网络的丙烯腈流化床反应器的模拟与离线优化

Simulation and Off-Line Optimization of an Acrylonitrile Fluidized Bed Reactor Based on Artificial Neural Networks

利用人工神经网络的方法建立了工业合成丙烯腈流化床反应器的数学模型。采用遗传算法与梯度下降法相结合的方法训练神经网络的权值和阀值。经过训练和可靠性检验的人工神经网络能够满足工业生产的模拟要求。利用单纯型算法与遗传算法相结合的优化方法对合成丙烯腈工业流化床反应器进行了操作条件优化 。

In this paper,a mathematical model is developed for an industrial acrylonitrile fluidized bed reactor based on artificial neural networks.In order to overcome a local optimal solution,a new algorithm,which combines the perfect characters of both GA(Genetic Algorithms)and GDR(Generalized Delta-Rule) is used to train ANN(Artificial Neural Networks).For the purpose of gaining the global optimal solution,a new algorithm called SM-GA,incorporating the advantages of both SM(Simplex Method) and GA,is proposed and applied to optimize the operating conditions of an acrylonitrile fluidized bed reactor in industry.\;This off-line simulation and optimization of an acrylonitrile fluidized bed reactor makes good bases for the real-time optimization control.