【作者】 郭红梅；

【导师】 于佐军；

【作者基本信息】 中国石油大学 ， 控制理论与控制工程， 2007， 硕士

【摘要】 工业生产过程广泛地存在着非线性现象，而线性预测控制难以满足非线性系统的控制要求。神经网络预测控制可以解决过程模型未知的复杂非线性系统的控制问题。所以本文研究了基于神经网络模型的预测控制的基本理论，并以酸碱中和反应的pH值控制为例实现神经网络预测控制仿真。在仿真过程中，借助Matlab和Simulink并结合S函数模块实现复杂的先进控制程序。首先研究了基于BP网、径向基函数网和小波网络的网络辨识模型，并通过仿真比较了这三种网络模型的精度和训练速度。接着采用梯度下降法求解非线性预测控制的优化问题，设计了基于小波网络模型和RBF网络模型的多步预测控制器。仿真结果表明，该控制算法实时性好，通过参数调节可使其达到较好的控制效果。虚拟仪器代表着仪器发展的最新方向和潮流，利用虚拟仪器实现工业过程的先进控制更具有一定的实际意义。所以本文还研究了如何基于虚拟仪器实现工业过程的实时先进控制。首先介绍了虚拟仪器技术、LabVIEW软件和数据采集的基本理论，然后利用LabVIEW与Matlab混合编程方法，对酸碱中和反应的pH值进行基于神经网络模型的预测控制仿真。最后采用数据采集卡、信号调理板、LabVIEW软件和温控炉对象，设计了温控炉的控制虚拟仪器。通过LabVIEW编程和界面设计，对模型未知的温控炉进行了基于神经网络模型的预测控制。实验结果表明，可以通过调节控制器参数使控制器达到较满意的控制效果。更多还原

【Abstract】 Nonlinear phenomena widely exist in the industry process and the linearpredictive control can’t qualify the control of the nonlinear system. Thepredictive control based on neural network model can control the complicatednonlinear systems whose models are unknown. So this paper studies the theoryof predictive control based on neural network (NN) model, and simulates thepredictive control based on neural network (NN) scheme taking example for thepH control of the neutralization process. Matlab, Simulink and the Sfunctionmodule in Simulink are used to design the complicated advanced controlprogram. Firstly, neural network identification models based on BP network,RBF network and wavelet network are researched. The simulation results provethat these predictive models using NN identification are efficient. Secondly, thegradient descent method is adopted to solve the optimal problem in thenonlinear predictive control and the multisteppredictive controllers aredesigned based on the wavelet net and RBF predictive models. The simulationresults show that this control algorithm has good real time characteristics. It cancontrol well by adjusting the parameters.Virtual instrument represents the development direction and fashion of theinstrument. It is significant to make use of virtual instrument to implementadvanced process control. So this paper also studies how to implement the real time advanced control in process based on the virtual instrument. After thetheories of the virtual instrument technology, the LabVIEW software and thedata acquisition are researched, the predictive control based on NN is simulatedusing the hybrid programming method with LabVIEW and Matlab, takingexample for the control process of the pH in the neutralization. In the end, avirtual instrument is designed using a DAQ board, a signal processing board,LabVIEW and a temperature control stove. Through programming anddesigning of the interface in LabVIEW, the predictive control based on NN iscarried out for the temperature control stove whose model is unknown. Theresults of the control experiment show that the controller can provide goodperformance by adjusting its parameters.更多还原