【作者】 李凯；

【导师】 刘朝英；

【作者基本信息】 河北科技大学 ， 检测技术与自动化装置， 2008， 硕士

【摘要】 集中供热是现代化城市的主要基础设施之一,安装热力表、实行分户计量已成为我国发展城市集中供热的必然趋势。采用计算机技术和先进的控制方法,对供热系统进行监测和控制,实施变流量运行,能够克服定流量供热系统存在的供热质量差、冷热不均、能源浪费、计费不合理、管理水平落后等缺点。针对变流量供热系统的时变性、耦合性和滞后性等特点,提出了换热站二次网温度控制和恒压点压力控制方案,并利用神经网络实现了二次网供、回水温度的解耦控制。静态解耦方法简单、易实现,但解耦不彻底,解耦效果时好时坏,造成系统运行不稳定。根据二次网供、回水温度间的耦合关系,提出了静态解耦和基于对角递归神经网络(DRNN)辩识的神经网络PID解耦相结合的控制方案,该方案利用静态解耦减弱变量间的耦合强度,利用基于DRNN的神经网络PID实现动态解耦,有效地克服了二次网供、回水温度间的耦合,同时将DRNN网络作为在线辩识器,自动调整网络权值,实现PID控制器参数的在线调整,克服了常规PID控制器参数难以整定和自适应能力差的缺陷。基于上述理论分析,采用静态解耦、常规PID解耦和基于DRNN的神经网络PID解耦的控制方案,对所建立的换热器数学模型分别进行了MATLAB仿真,仿真结果表明,采用带静态解耦的神经网络PID解耦控制方案,实现了动态近似解耦,静态完全解耦,具有响应速度快、超调量小的品质。在MATLAB仿真基础上,利用PCS-B型过程控制系统实验装置进行了实验研究,实验结果表明,采用神经网络PID控制器,动态调节时间和稳态误差明显减小,获得更加满意的控制效果。按照“集中管理、分散控制”的思想,设计了以工控机为上位机、PLC为下位机,组态软件MCGS为操作平台的城市集中供热监控系统。在MCGS环境下,完成了供热系统监控中心的主画面、神经网络PID控制画面、操作面板画面、数据显示和历史曲线显示等界面的设计。该方案的应用可以提高供热企业的管理水平和经济效益。更多还原

【Abstract】 Central heating is one of the main infrastructures of modern cities. The installation of calorimeter, achieving household measurement, has become an inevitable trend of development for our urban central heating. Using computer technology and advanced control method, monitoring and controlling heating system, implementing variable flow operation, can overcome the shortcomings of poor quality, hot and cold inhomogeneous, wasted energy, unreasonable fees and management level backward in fixed flow heating system in past. Against(consider) the characteristics, of coupling, and lagging of variable flow heating system, secondary network temperature control programs and constant pressure point pressure control program are designed, and decoupling control is realized by using neural network.Although the static decoupling method is simple, easy to realize, the decoupling can not be made thoroughly which can make the system unsteady. According to the coupling relationship of temperature of water supply and return, a decoupling method that static decoupling combines with PID decoupling based on diagonal recurrent neural network (DRNN) is proposed. In the method, the intensity of coupling is reduced by using static decoupling and dynamic decoupling is realized by using PID decoupling based on DRNN, to overcome coupling of temperature of secondary water supply and return. Meanwhile, consider the defect of that parameter regulation is difficult and bad adaptive capacity for conventional PID controller, neural network PID controller based on DRNN is designed. DRNN network is regarded as online identification and can be adjusted the right value of the network automatically, to realize PID controller parameter adjustment online. Based on analysis of theories above, adopting the methods of static decoupling, conventional PID decoupling and neural network PID decoupling, MATLAB simulations are carried for mathematical model of heat exchanger established. MATLAB simulation results shows that using neural network PID decoupling control programs with static decoupling has performed dynamic approximate decoupled and static completely decoupled, with the quality of fast response and small overshoot. Furthermore, Using PCS-B process control system experimental device, experimental study has been done. The results show that neural network PID controller can improve quality of dynamic process and reduce static error in practical application. And the satisfactory effect can be reached.According to "centralized management, decentralized control" thought, urban central heating system adopts three-tier structure. And the program of heating station monitoring control system is designed, IPC is upper computer and PLC is subordinate computer based on MCGS configuration software platform in this program. In MCGS environment, main screen, neural network PID control screen, operation panel screen and data curve are designed for monitoring center in heating system. The application of this program would improve remarkably management level and economy benefit for heating enterprise.更多还原