毕业设计_神经网络控算法仿真.doc

摘 要 目前，由于PID结构简单，可通过调节比例积分和微分取得基本满意的控制性能，广泛应用在电厂的各种控制过程中。电厂主汽温被控对象是一个大惯性、大迟延、非线性且对象变化的系统，常规汽温控制系统为串级PID控制或导前微分控制,当机组稳定运行时,一般能将主汽温控制在允许的范围内。但当运行工况发生较大变化时,却很难保证控制品质。因此本文研究基于BP神经网络的PID控制，利用神经网络的自学习、非线性和不依赖模型等特性实现PID参数的在线自整定，充分利用PID和神经网络的优点。本处用一个多层前向神经网络，采用反向传播算法，依据控制要求实时输出Kp、Ki、Kd，依次作为PID控制器的实时参数，代替传统PID参数靠经验的人工整定和工程整定，以达到对大迟延主气温系统的良好控制。对这样一个系统在MATLAB平台上进行仿真研究，仿真结果表明基于BP神经网络的自整定PID控制具有良好的自适应能力和自学习能力，对大迟延和变对象的系统可取得良好的控制效果。 关键词：主汽温，PID，BP神经网络，MATLAB仿真 ABSTRACT At present, because PID has a simple structure and can be adjusted proportional 、integral and differential to satisfactory control performance, it is widely used in power plants of various control process. The system of power plant main steam temperature is an large inertia、big time-delayed and nonlinear dynamic system. Conventional steam temperature control system adopted cascade PID control or the differential control of lead before. When the unit is stable, these methods will control the steam temperature in a certain range ,but when operating conditions changed greatly, it is difficult to ensure the quality of control. This article studies PID control based BP neural network . Using such characteristics of neural network self-learning, nonlinear and dont rely on model realize PID parameters auto-tuning. It can make full use of the advantages of PID and neural network. Here, we use a multilayer feedforward neural network using back propagation algorithm. This net can real-time output Kp, Ki, Kd as the PID controller parameters , insteading of the traditional PID parameters determined by experience, so it can obtain good control performance .For such a system ,we can simulate in MATLAB simulation platform. The simulation results show that the PID control based BP neural network has good adaptive ability and self-learning ability. For the system of large delay and free-model can obtain good control effect. KEY WORDS: main steam temperature