提出一种利用回声状态网络(echo state network,ESN)建立复杂分布参数系统模型的灰箱建模方法。此建模方法可以充分利用已知机理模型的结构信息和回声状态网络的逼近能力,可更好地描述和解释出系统各变量之间的因果关系,使模型的"灰箱"化程度更高。首先,根据系统方程和先验知识将初始系统特征团引入ESN储备池中,赋予网络节点实际物理意义,并以此建立结构逼近神经网络模型;然后,通过逐步回归分析方法,结合递归最小二乘算法选择最优系统特征团,并对网络结构进行优化,建立起描述系统特性关系的灰箱模型。本文以实验室规模的管式聚合反应过程作为实验对象,建立以温度分布为输出的数学模型,结果表明所提出的灰箱建模方法行之有效。
A self-organizing radial basis function(RBF) neural network(SODM-RBFNN) was presented for predicting the production yields and operating optimization. Gradient descent algorithm was used to optimize the widths of RBF neural network with the initial parameters obtained by k-means learning method. During the iteration procedure of the algorithm, the centers of the neural network were optimized by using the gradient method with these optimized width values. The computational efficiency was maintained by using the multi-threading technique. SODM-RBFNN consists of two RBF neural network models: one is a running model used to predict the product yields of fluid catalytic cracking unit(FCCU) and optimize its operating parameters; the other is a learning model applied to construct or correct a RBF neural network. The running model can be updated by the learning model according to an accuracy criterion. The simulation results of a five-lump kinetic model exhibit its accuracy and generalization capabilities, and practical application in FCCU illustrates its effectiveness.
The paper describes a closed-loop system identification procedure for hybrid continuous-time Box–Jenkins models and demonstrates how it can be used for IMC based PID controller tuning. An instrumental variable algorithm is used to identify hybrid continuous-time transfer function models of the Box–Jenkins form from discretetime prefiltered data, where the process model is a continuous-time transfer function, while the noise is represented as a discrete-time ARMA process. A novel penalized maximum-likelihood approach is used for estimating the discrete-time ARMA process and a circulatory noise elimination identification method is employed to estimate process model. The input–output data of a process are affected by additive circulatory noise in a closedloop. The noise-free input–output data of the process are obtained using the proposed method by removing these circulatory noise components. The process model can be achieved by using instrumental variable estimation method with prefiltered noise-free input–output data. The performance of the proposed hybrid parameter estimation scheme is evaluated by the Monte Carlo simulation analysis. Simulation results illustrate the efficacy of the proposed procedure. The methodology has been successfully applied in tuning of IMC based flow controller and a practical application demonstrates the applicability of the algorithm.
